feat(api): add Buddy Showcase focus filtering

[Mygod]

Summary

• add an exact Buddy Showcase contest_focus selector to Pokéstop and combined fort-scan DNF filters
• apply the structured focus predicate before max_fort_results, so unrelated Showcases cannot consume the endpoint cap
• expose active showcase_focus values as native JSON objects and advertise showcase_focus_filter: true through both availability endpoints
• load showcase_focus during cache-miss/startup population and keep only a compact Buddy-level projection in the hot fort lookup

Example filter:

{
  "contest_focus": [
    {
      "type": "buddy",
      "min_level": 3
    }
  ]
}

Compatibility

• existing contest_pokemon and contest_pokemon_type filters are unchanged
• existing availability pokemon_id, form, and type_id mirrors remain available
• legacy rows without structured focus return showcase_focus: null
• the existing per-stop showcase_focus wire representation is unchanged
• clients can treat a missing or false showcase_focus_filter capability as an older Golbat

Related to WatWowMap/ReactMap#1090.

Verification

• go test -tags go_json ./...
• go build -tags go_json ./...
• focused go test -race ./decoder/ coverage for parsing, indexing, DNF matching, cap ordering, and availability
• go run github.com/golangci/golangci-lint/v2/cmd/golangci-lint@v2.12.2 run — 0 issues
• git diff --check origin/main...HEAD

[jfberry]

Thanks for this — I've cherry-picked 8e10f6c into #393 (feat/fort-api-status-and-limits) as ea44536, with your authorship preserved, so the showcase filtering lands together with the availability enrichment it sits next to. #393 touches the same three structs, so keeping them on one branch avoids a merge race between the two PRs.

Conflicts were all additive and resolved by keeping both sides:

• FortLookup carries ShowcaseBuddyMinLevel alongside feat(api): status endpoint, fort scan limit status, availability enrichment #393's ShowcaseRankingStandard. Both int8s share the existing pad before ShowcaseExpiry, so the struct is still 184 bytes (verified with unsafe.Sizeof) — your projection stays free.
• showcaseKey now dedupes on ranking standard as well as structured focus.
• ApiPokestopShowcaseAvailable returns both ranking_standard and showcase_focus.

Two small changes on top:

1. The showcase_focus nullability doc now reads "null for legacy rows without showcase_focus and for rows whose stored focus could not be parsed" — updatePokestopLookup logs and falls back to nil when parseShowcaseFocus fails, so a modern row with unparseable JSON also surfaces null.
2. contest_ranking_standard and raid_temp_evolution_id DNF filters were added, following the pattern your contest_focus established: your PR was the one that correctly wired a new dimension through lookup, availability index and filter in the same change, which is what made the two missing filters on the feat(api): status endpoint, fort scan limit status, availability enrichment #393 side visible.

One thing worth flagging that this PR inherits rather than introduces. ContestProto.Focuses is a repeated field, and updatePokestopFromGetContestDataOutProto builds focusStore as a Go map, then calls SetShowcaseFocus once per iteration — so when a showcase carries more than one focus, only the last-visited entry survives, non-deterministically per decode (there's already a SHOWCASE: we got more than one showcase focus warn guarding that case). That was harmless while showcase_focus was informational, but it's now load-bearing for ShowcaseBuddyMinLevel and the contest_focus filter, so a dual-focus showcase could flicker in and out of matching between GMOs. Not a blocker for this change, and the fix belongs in the decode path — deriving the Buddy projection by scanning focusStore for a buddy entry specifically, rather than depending on whichever key iteration landed last.

Closing in favour of #393 once that merges — the commit and the credit carry over.