refactor: externalize section keyword config to sections.config.json

[s-annam]

Summary

Moves the three hard-coded section-keyword structures out of `regex.ts` into a
single editable JSON file (`sections.config.json`) with a typed loader
(`sections.config.ts`). Pure refactor — no behaviour change. Same aliases, same
`SectionName` union, same exports to all existing consumers.

• `sections.config.json` — canonical source for 8 sections: aliases,
  anchors (L2-ready), `splitLetterNormalizable`, and `anchorFallback`. Version
  field `"1"` for future migrations.
• `sections.config.ts` — typed loader: imports JSON, validates shape at
  module load, derives `SECTION_KEYWORDS` + `SPLIT_LETTER_NORMALIZABLE_SECTIONS`
  (unchanged public API) + new `SECTION_ANCHORS`/`SECTION_ANCHOR_FALLBACK`
  accessors (unread until L2/L3). Drift guard: `const _drift: Record<SectionName,
  SectionConfig> = config.sections` fails the build if a JSON key is missing from
  the union.
• `regex.ts` — deletes the three inline structures; imports + re-exports them
  from `sections.config.ts` so `sections.ts`, `markdown-lines.ts`, and
  `extract-fields.ts` import paths are unchanged (zero edits to those files).
• `tsconfig.app.json` — adds `resolveJsonModule: true` (required for tsc;
  Vite already handled JSON at runtime).

Closes #110

Test plan

• npm run typecheck clean (34 files, 0 errors)
• npm run test green (34 test files, 414 tests, 0 failures)
• Drift guard verified: renaming "summary" key in JSON to "summaryX" produces TS2741 at compile time; reverted
• Manually verified in npm run dev / npm run preview

[rohithgollapalli]

Review — Approve ✅

Clean, behaviorally faithful refactor. No correctness bugs found.

Correctness — verified equivalent

Checked the two derived structures against the deleted originals:

• SECTION_KEYWORDS — all 8 sections, identical alias lists, identical key order (summary → experience → … → other). JSON preserves insertion order and Object.fromEntries(Object.entries()) preserves it, so iteration order is unchanged.
• SPLIT_LETTER_NORMALIZABLE_SECTIONS — the JSON splitLetterNormalizable: true set resolves to {summary, experience, education, projects, certifications}, exactly the original hand-written set (skills correctly stays false, preserving the Letter-spaced section headers ('E XPERIENCE') parse with zero experiences, disabling per-bullet grouping #56 sidebar-bleed guard).

Order-dependence doesn't matter regardless — every alias is globally unique across sections, so both matchSectionHeader (.includes) and normalizeSplitLetterHeaders (keywordToSection map) produce identical results.

Type/build safety — sound

• Both consumers (regex.ts, markdown-lines.ts) read SECTION_KEYWORDS through the explicit as Array<[SectionName, readonly string[]]> cast, so dropping as const for Record<SectionName, readonly string[]> is compatible.
• .ts import extension matches the repo-wide convention; allowImportingTsExtensions: true is already set.
• import rawConfig from "*.json" is correctly gated by the newly-added resolveJsonModule: true. Import/re-export-from-same-module in regex.ts is valid TS. No circular import.
• Removing the unused SECTION_ANCHORS/SECTION_ANCHOR_FALLBACK accessors was the right call — keeps fallow green and defers the export to the L2/L3 PR that adds the reader.

Minor (non-blocking)

1. Drift guard is one-directional. const _drift: Record<SectionName, SectionConfig> = rawConfig.sections catches a renamed/deleted key, but an extra JSON key is allowed by structural assignment and would flow into SECTION_KEYWORDS at runtime mislabeled as a SectionName. Low practical risk; worth a one-line comment noting the guard only covers missing keys.
2. Runtime validate() is partial vs. compile-time — validates aliases/anchors arrays but not the splitLetterNormalizable/anchorFallback booleans (those are enforced at compile time via the _drift assignment). Consistent and fine, just slightly asymmetric.
3. JSON anchors/anchorFallback are validated/typed but unread until L2 — Head-noun anchor fallback for qualified section headers (closes #108) #111/L3 — Visual-primary boundary detection + name/contact disambiguation #112 — data, not dead exports, as the PR notes.

Before merge

The test-plan manual dev/preview box is unchecked. Given full equivalence + 414 tests green (incl. regex.test.ts + split-letter-header.test.ts which pin section matching), I don't consider it blocking — but ticking it costs nothing.

Solid, well-scoped refactor that sets up the L2/L3 anchor work cleanly.

[s-annam]

Merged — thanks for the thorough review, @rohithgollapalli. Confirmed the equivalence checks and the #56 sidebar-bleed guard call. Your three minors were all non-blocking so I merged on your approval (dismiss-on-push would've bounced it); filed the drift-guard comment correction (your minor #1) as a follow-up rather than re-trigger a review cycle.