Show the score in the graphical UI, and scope format.sh to skip vendored code

[dmccoystephenson]

Summary

• Formatting is scoped and up to date (format.sh reformats 9 unrelated files, and would rewrite the vendored src/lib #120). black's exclusion of the vendored src/lib now lives in a new pyproject.toml [tool.black] section, and autoflake is given --exclude lib, so re-vendoring an upstream copy of graphik or py_env_lib is no longer turned into a merge conflict. The eight non-vendored files that had drifted out of black's style are brought back into line in one pass, so subsequent diffs stay scoped to their actual change. Landing that first is also why the second commit's diff is readable.
• The score is drawn in the graphical UI (Graphical UI never displays the score, while the text UI shows it live #124). drawHud() now leads with the score alongside the currency readout it already drew; previously the score was reachable only through the text UI, the end-of-run console summary and the save record, which left the score-multiplier annotation with no line to attach to in the default UI. The score shares the first HUD line with the currency because graphik centres each string on the x it is handed, so two centred strings on one row would overlap.
• The multiplier annotation rule is shared. Whether and how (x2) is appended moves into scoring.formatScoreLabel, which both the graphical HUD and TextRenderer.renderStats read, so the two UIs cannot restate it differently — the drift that PRs Add text-based UI option for terminal gameplay #92, Add text-based UI option with clean architecture, comprehensive testing, CI/CD verification, and performance optimizations #95 and Add cross-run progression: obituaries, cosmetics, currency/shop, naming/lore, ascension #99 were each about. formatScoreLabel is a pure function alongside the scoring rules the renderers are already handed plain numbers from, so no new gameplay-to-UI coupling is introduced.
• The README's scoring section is corrected on the one point that changed: the (x2) annotation is no longer text-UI only.

Not included: #122

Issue #122 was investigated and is not reproducible as described; it has been left open with the evidence posted as a comment there, because deciding what should happen instead is a design call rather than a bug fix. In short: checkForLevelProgressAndReinitialize() is only ever reached from restartRun() and the collision branch of moveEntity(), both of which call recordCurrentRun() first, so a run is always exactly one level and the score is never discarded mid-run.

Test plan

• python3 -m pytest — 223 passed (217 on main)
• python3 -m compileall src — clean
• python3 -m black --check src tests — clean, and src/lib is left untouched
• New coverage: formatScoreLabel at a neutral, doubled and fractional multiplier, and the graphical HUD's score line both with and without a multiplier running
• README controls table and usage commands re-checked against handleKeyDownEvent and the --text-ui argparse flag

Closes #120
Closes #124

This PR description was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

[dmccoystephenson]

Self-review

The full diff was read against this repository's conventions. Nothing blocking was found. Three judgment calls are worth a reviewer's attention, and one finding was applied before this comment was posted.

Applied

tests/rendering/test_hud_and_banner.py:41 — the drawText stub captured the drawn string via a lambda that spelled out every one of graphik's five positional parameters, which would break on any caller that passed one by keyword. It now takes (text, *args), matching the stub already used in tests/ui/test_banner.py:70. Fixed in d301e0f.

For a reviewer to confirm

src/textui/textrenderer.py:8 — this adds the text renderer's first import of a non-stdlib project module (from scoring.scoring import formatScoreLabel). The decoupling asked for in PR #95 was read as "gameplay must not reach into a UI package, and a UI package must not reach into gameplay state"; scoring/scoring.py holds pure functions over plain numbers with no game object, no state and no UI dependency, the same shape as progression/obituary.py's formatObituaryScreen, which gameplay already imports for display lines. The alternative — restating the (x2) rule inside drawHud as well — is the duplication that PRs #92, #95 and #99 were each cleaning up, so sharing the function was preferred. Should the boundary be read more strictly than that, the fallback is for renderStats to accept an already-formatted label from Ophidian.getScoreLabel(), at the cost of a signature change to a method three existing tests call by keyword.

src/scoring/scoring.py:62 — the annotation triggers on multiplier > 1, so a hypothetical multiplier below 1 would render unannotated. This is the pre-existing renderStats behaviour moved verbatim rather than a new decision, and no such power-up exists today (getScoreMultiplier returns 1.0 or 2.0), so it was left alone rather than widened speculatively.

src/ophidian.py:322 — the score shares the first HUD line with the currency (Score: 125 | Currency: 42) because graphik.drawText centres each string on the x it is handed, so two centred strings on one row would overlap. A long score plus a long currency balance could therefore run wide on a narrow window. The upgrades line directly below already has that property, so no new mechanism was introduced for it; issue #72 is where a dedicated HUD area is being tracked.

Checked and clean

• format.sh's exclusion was verified to take effect, not just to be declared: black --check src tests reports 9 files before the change and 8 after, with src/lib/pyenvlib/environment.py dropping out, and src/lib is unmodified in the diff.
• The formatting-only commit was confirmed to be formatting-only — it is separated from the functional commit precisely so that can be read at a glance.
• The # /bin/bash line added to format.sh is a comment rather than a shebang, matching test.sh exactly; both scripts are run as ./x.sh and fall back to the invoking shell, so behaviour is unchanged.
• The README controls table was re-checked key by key against handleKeyDownEvent (w/a/s/d and arrows, F11 graphical-only, l, c, p, r, q), and the usage commands against the --text-ui argparse flag. Only the multiplier-annotation sentence needed correcting.
• python3 -m pytest: 223 passed, against 217 on main.

This comment was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).