feat: add my-taste builtin profile (personalization-forward feed)

[ham-33]

What

Adds a new builtin feed profile, my-taste, tuned for viewers who want the feed to lean toward their own signals rather than global popularity.

Knob	for-you (default)	my-taste
engagement weight	0.35	0.15
affinity weight	0.30	0.40
topic weight	0.20	0.30
social_proof weight	0.15	0.10
trending weight	0.10	0.05
freshness tau	8 h	10 h
in-network ratio	0.5	0.6
explore slot	every 10th	every 8th
prolific damp threshold	10	8

The net effect: accounts you engage with and subjects you care about lead, while off-topic viral posts sink. The diversity guards (author cap, topic-run break, exploration slots, language/seen/stranger-reply penalties) are left on, so it sharpens personalization without collapsing into a filter bubble.

Why

This profile came out of a hands-on comparison of for-you vs. a personalization-forward weight set on a shared candidate set. Relative to the default, my-taste pulls followed / on-topic content higher, drops off-topic viral posts out of the page, and surfaces slightly more new-author discovery — while the same seen / language / prolific / stranger-reply filters keep the feed from feeling "off".

Changes

• profiles.go: add MyTasteConfig() and register "my-taste" in BuiltinProfiles().
• profiles_test.go: TestMyTasteProfilePersonalizesOverPopularity asserts the profile prefers a followed, on-topic post over an off-topic viral one more strongly than the default, and that the personal post wins outright under my-taste.
• README.md, ALGORITHM.md: add the profile to the feed-profiles tables.
• CHANGELOG.md: Unreleased entry.

No changes to the scoring formula, sourcing, filters, or page rules — this is purely an additional published weight set.

Testing

go vet ./...            # clean
gofmt -l .              # clean
go test -race -cover ./...   # ok, 98.8% coverage

🤖 Generated with Claude Code

https://claude.ai/code/session_014kaXMnwzpviZesDmpphxNn

---

Generated by Claude Code