feat(calc): implement DistanceRamp tag (Close/Far Combat distance MORE)

[ackness]

Summary

Implements the PoB2 DistanceRamp modifier tag — the distance-interpolated MORE/INC
damage that Close Combat / Far Combat / Point Blank / Far Shot grant. Previously
translate_tag had no DistanceRamp arm, so the whole stat-map entry fell to
UnsupportedTag and was silently dropped.

This PR also lands the oracle-verified correction of the offensive Close-Combat
cluster roadmap, which turned out to be a static-analysis misdiagnosis.

What changed (commit d3456ec)

• New ModTag::DistanceRamp { ramp } + linear interpolation with endpoint clamp in
  Modifier::effective_number (ramp_factor, vendor ModStore.lua:574-590 line-for-line).
• translate_tag parses the ramp point list (stat_map_engine.rs); matches() and
  canonical_tag() get the new arm.
• skillDist source = new CalcConfig::skill_distance, set in the orchestrator to
  mode_effective && the enemyDistance <Input> value ONLY. This mirrors vendor exactly:
  skillCfg.skillDist = env.mode_effective and env.configInput.enemyDistance
  (CalcActiveSkill.lua:655), where configInput is the <Input> channel and does
  not include <Placeholder> values. The placeholder-inclusive
  Multiplier:enemyDistance (ConfigTab apply fallback) is a separate channel feeding
  only the accuracy distance penalty.

Why it is dormant for the demo suite (and that is correct)

All 18 demo-bd-test builds carry enemyDistance as <Placeholder number="20">, none as
<Input> → skill_distance == None → every DistanceRamp mod is skipped, matching the
golden output (PoB2 likewise applies no Close Combat distance MORE to these builds).
parity_no_regression is therefore unchanged. The tag activates only for builds that set
enemyDistance as an explicit <Input>.

Oracle proof (flicker-strike): editing the build's enemyDistance 5 → 40 leaves AvgDamage
unchanged at 108623.8 (== golden); the MoreDamage intermediate is 1.3 = Concentrated Area
only, no Close Combat. The roadmap's expectation that DistanceRamp would move
monk-twister / flicker / shield-wall parity is void.

Roadmap correction (commit 7fabaca)

devs/docs/audits/offensive-parity-roadmap-2026-06-28.md gets a "MAJOR CORRECTION" section
recording the oracle-verified facts: #2 is not a parity-mover (placeholder enemyDistance),
monk-twister has no Close Combat on its Twister group, and #1 Onslaught is a real
over-application but the roadmap's Condition:UsingFlask fix does not work (the grantor is a
trigger-gated Silver Charm). Methodology lesson: oracle-verify every offensive root cause
before implementing.

Test plan

• cargo fmt --check clean
• cargo clippy --workspace --all-targets -- -D warnings clean
• cargo nextest run --workspace — 2097 passed, 5 skipped
• cargo test -p pobr-build --test parity — parity_no_regression unchanged (dormant tag)
• New unit tests: interpolation at distance 20 (→ 0.6 factor), endpoint clamp, and the
  None-skip path (panel / placeholder-only enemyDistance)