Replace scan extraction with @dittowords/text-extract

[jholiga]

Overview

The CLI held its own copy of the extraction pipeline. That code now ships as @dittowords/text-extract. This PR imports the library and deletes the local copy.

• Add @dittowords/text-extract@^0.2.0 as a dependency.
• Delete lib/src/scan/{extract,walk,rules,types}.ts and all of lib/src/scan/lang/.
• Delete the 19 test files for the deleted code. The library tests it.
• Import DittoScanCandidate, DittoScanExtractSummary and runExtract from the library in commands/scan.ts, http/scan.ts and ditto.ts.
• Remove @ast-grep/napi, @ast-grep/lang-kotlin, @ast-grep/lang-swift, globby and yaml from package.json. The library owns them now.
• Log the library's new summary.filesFailed count and the failure detail. A failed file drops out of the results, so a non-zero count means strings are missing.
• Add lib/src/scan/extract.test.ts. It checks that the CLI resolves the library and gets usable candidates from testfiles/.

lib/src/scan/analyzeDirectories.ts stays. It is CLI-only code and the library does not have it.

Net: -5,318 lines.

Context

The library is a superset of the deleted code. It adds a single-file entry point (extractFile) and a failure count in the summary. The CLI only uses runExtract, and that behaves the same.

The .d.ts bundle now re-exports the scan types from the library instead of inlining them. @dittowords/text-extract is a declared runtime dependency of the published package, so consumers resolve it.

Screenshots

No visual changes.

Test Plan

Testing successfully completed locally via:

• yarn test — 39 suites, 285 tests pass.
• npx tsc --noEmit --module preserve — no errors. (Plain npx tsc --noEmit fails on a pre-existing tsconfig.json module/moduleResolution mismatch, on master too.)
• node esbuild.mjs — the bundle and bin/ditto.d.ts build.
• ditto scan testfiles --local — the candidate set is byte-identical to master's, after sorting.
• ditto scan lib --local — the candidate set was byte-identical to master's before the source files were deleted.

Out of scope

ignore and @google/genai are unused dependencies on master already. This PR leaves them.

🤖 Generated with Claude Code