Fix AA-identity/property-value mismatch in sequenceEncoder()/sequenceDecoder() property mode

[Qile0317]

Bug

sequenceEncoder(mode="property", ...) and sequenceDecoder(mode="property", ...)
silently mismatch amino-acid identity to property values for 3 of the ~10
supported property sets: crucianiProperties, MSWHIM, ProtFP.

Both functions assume .aa.property.matrix()'s output is ordered by the
canonical amino-acid alphabet, positionally — sequenceEncoder()'s roxygen
docs state this explicitly ("rows correspond to the 20 canonical amino acids
in the order of sequence.dictionary"), and sequenceEncoder()'s C++
backend / sequenceDecoder()'s .propertyDecoder() both index into it
purely positionally, never by name.

.aa.property.matrix() didn't actually guarantee that. It returns
Peptides::AAdata[[key]] reshaped via do.call(rbind, ...), in whatever
column order that scale's author originally stored it in Peptides — which is
NOT canonical for these 3 sets:

property set	actual stored order
crucianiProperties	canonical except positions 6,7 (E,Q instead of Q,E)
MSWHIM	fully alphabetical
ProtFP	fully alphabetical

calculateProperty() already reorders by name after calling this helper
(S[, amino.acids, drop=FALSE]); sequenceEncoder()/sequenceDecoder()
didn't, so e.g. sequenceEncoder("R", mode="property", property.set="MSWHIM", max.length=1)$flattened silently returned Cysteine's true MSWHIM values, not
Arginine's.

Verified against Peptides::mswhimScores()/crucianiProperties()/protFP()
— Peptides' own independent public scoring functions, not just immApex/
calculateProperty() internals — confirming the canonical-order reading is
the objectively correct one, not just a different convention.

Fix

Centralizes the canonical-AA reorder inside .aa.property.matrix() itself
(both the .builtin_scales branch and the Peptides::AAdata branch), so
every caller gets canonical order by construction instead of relying on each
call site to remember to reorder. This fixes both bug sites in one change:
R/sequenceEncoder.R:132 and R/sequenceDecoder.R:110 (the latter has the
identical root cause but wasn't previously flagged).

calculateProperty()'s own existing reorder is left in place — now
redundant but harmless/idempotent.

Tests

• .aa.property.matrix() returns canonical column order for every supported
  property set (test-calculateProperty.R).
• sequenceEncoder() property mode matches Peptides' own scoring functions
  for single-residue probes on all 3 previously-broken sets
  (test-sequenceEncoder.R).
• sequenceDecoder() property mode correctly identifies an amino acid from a
  cube built directly from Peptides ground truth — not via
  sequenceEncoder(), since a pure round trip through immApex's own
  (formerly buggy) functions would self-cancel this exact bug
  (test-sequenceDecoder.R).

Version

1.7.1 → 1.7.2, NEWS.md updated.

[Qile0317]

@ncborcherding Note: the test-coverage fail related to buildNetwork seems to be a separate issue that's probably to do with some obscure C++ behaviour across different versions or checker/compiler params, etc.

[codecov]

Codecov Report

❌ Patch coverage is 85.00000% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
R/calculateProperty.R	87.50%	2 Missing ⚠️
R/sequenceDecoder.R	66.66%	1 Missing ⚠️

Files with missing lines	Coverage Δ
R/sequenceEncoder.R	97.64% <100.00%> (ø)
R/utils.R	55.03% <ø> (ø)
R/sequenceDecoder.R	94.20% <66.66%> (ø)
R/calculateProperty.R	88.31% <87.50%> (-0.76%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.