Rename PhaseAmbiguousEffect -> PhaseCandidateSet

[iskandr]

Summary

Renames the public class returned by germline-aware effect prediction when somatic + germline share a codon and phase is unknown:

• Before: PhaseAmbiguousEffect (singular, describes the cause).
• After: PhaseCandidateSet (plural-flavored, describes the contents — a set of phase-hypothesis candidates).

The class wraps .candidates (a tuple of per-hypothesis classified effects) and exposes .outcomes (same candidates plus per-haplotype evidence metadata). The plural name reads better at call sites:

if isinstance(eff, PhaseCandidateSet):
    for c in eff.candidates:
        ...

This is a hard rename — no deprecation alias. The class was only accessible via varcode.effects.effect_classes.PhaseCandidateSet (never re-exported from varcode/__init__.py), so the blast radius is limited to external code that imported the old name directly.

Bumps to 4.23.0 (breaking).

Docs rewrite

docs/germline.md was rewritten to be substantially clearer and to demonstrate the resolver-injection pattern with concrete output. It now leads with three side-by-side scenarios on the same CFTR test-fixture coordinates (7:117531100 T>A somatic + 7:117531101 T>C germline, same codon):

1. No germline (default) → Substitution p.L159M.
2. Germline + unknown phase → PhaseCandidateSet ?p.L159M, with .outcomes listing both haplotypes (B: ref-relative L->M; A: cis-germline S->T).
3. Germline + phase_resolver forcing cis or trans → single Substitution (p.S159T for cis, p.L159M for trans).

This makes the "varcode enumerates, an external phase source collapses" design explicit instead of buried prose.

Files touched

• varcode/effects/effect_classes.py — class rename + comment update.
• varcode/germline.py — three internal references.
• varcode/variant.py, varcode/effects/effect_prediction.py — docstrings.
• tests/test_germline_effects.py — class import + isinstance assertions (8 references).
• docs/germline.md — full rewrite with the three-scenario demo.
• varcode/version.py — 4.22.1 -> 4.23.0.

Test plan

• pytest tests/test_germline_effects.py — 34/34.
• pytest tests/ — 1185/1187 (2 skipped, same as before).
• Manual: ran the doc's three scenarios in a REPL and confirmed the output matches the comments verbatim.

Out of scope

• Did not touch .candidates / .outcomes attribute names. .outcomes is part of the MultiOutcomeEffect contract shared with SpliceOutcomeSet and SV machinery; renaming there would cascade.
• No deprecation alias. If anyone files an issue, we can add PhaseAmbiguousEffect = PhaseCandidateSet in a follow-up patch.