fix(explore): surface camel-infix field definers on service-layer codebases (#1196)

[umi008]

Summary

Multi-word field-name queries (e.g. profileInfo isTrialEligible quotaInfo billingMethod) used to return 134 unrelated symbols and
bury the defining controller and service files under a dense neighbor.
Three compounding retrieval bugs were dead on JS/TS/Ruby/Python
service-layer codebases, where the camel-infix definer of a business
field is a method, not a class.

Root causes

1. context builder 5b (CamelCase-boundary LIKE) used
   name.indexOf(titleCased) after titleCased.toLowerCase(), so
   a multi-hump token ("Profileinfo") could never indexOf into
   "getProfileInfoV2". The SQL layer matched case-insensitively; the
   JS boundary check silently re-tightened that and dropped the row.

2. context builder 5b AND 5c restricted the LIKE batch to declaration
   kinds (class/interface/struct/trait/protocol/enum/type_alias), so on
   a method-centric codebase they contributed nothing. Single-word
   tokens like "search" → "Search" were unaffected because they
   don't go through the LIKE path the same way; this is why the bug
   wasn't caught during tuning.

3. explore's named-symbol seeding was exact-name only. A field-name
   token (profileInfo) has no node of its own, so it contributed
   zero seeds: no +50 file score, no entryFiles protection, no
   named-file sort. The only surviving entry points were weak FTS
   sub-token hits (camelCase split gives quota, payment, …) and
   a depth-3 BFS from those dragged in a huge neighbor whose body
   filled the render budget.

FTS can't cover this case by design: camelCase identifiers are single
FTS tokens, so "profileinfo"* never matches an interior hump —
that's what 5b/5c were built for, and they were disabled by (1)+(2).

Fix

• 5b: case-insensitive boundary check, with a guard that the
  boundary char itself is uppercase (so we still require a real hump).
• 5b/5c: a separate LIKE batch for callable kinds
  (function/method/component), merged before the boundary
  filter. Per-batch 200-row cap so a hot single-word term can't crowd
  classes out of the length-ordered batch.
• explore named-symbol seeding: new
  findCamelInfixCallables(cg, token) — a LIKE-based fallback that
  accepts prefix or camel-hump boundary matches on method/function/
  component kinds, case-insensitive. Fires only when the exact
  lookup yields no callable, so well-named queries are unaffected.
  Exposed CodeGraph.findNodesByNameSubstring publicly so this can
  be reached through the public API.

Validation

• New test suite __tests__/context-multi-word-fields.test.ts
  builds a mini service-layer JS repo (controller/profileController.js
  with getProfileInfo/getProfileInfoV2, service/billing.js with
  _getCustomerBillingMethods) and asserts:
  • 5b recovers getProfileInfoV2 from a profileInfo token
  • callable kinds participate, so 5b surfaces _getCustomerBillingMethods
    for a billingMethod token
  • the multi-word field query from the issue surfaces BOTH defining
    methods in the same call
  • the camel-hump boundary check is precise: profileinfo matches
    getProfileInfoByName (true hump) but NOT reprofileinfoBogus
    (mid-word lowercase run)
• All 4 new tests pass on Linux + macOS.
• Existing context-ranking.test.ts (4), context.test.ts (17),
  explore-result-count.test.ts, explore-output-budget.test.ts,
  adaptive-explore-sizing.test.ts, explore-corroboration-ranking.test.ts,
  explore-blast-radius.test.ts — 63 tests — all still pass.

Reproduction

query: "profileInfo isTrialEligible quotaInfo billingMethod"

Before: 134 symbols across 26 files; defining files
controller/profileController.js and service/billing.js absent.
After: those two files rank at the top.

Closes #1196