chore(deadcode): ratchet down the 85 pre-existing unused exports/types (follow-up to #607)

Follow-up to #607 (PR #673).

After #673 re-greened the dead-code ratchet, the knip baseline holds at **85** — all of it pre-existing drift, no longer masked by the devDep/false-positive noise:

| category | count |
|---|---|
| \`exports\` | 44 |
| \`types\` | 41 |

## Why this matters

The ratchet job (\`check:deadcode-ratchet\`) is **advisory / non-blocking** today (\`continue-on-error: true\`). Its stated milestone is to *"flip to blocking once the baseline is driven to zero."* Because it is non-blocking, roughly 7 genuine unused export/type regressions drifted into \`main\` unnoticed inside the +13 that #607 reconciled. Every merge that adds dead code under a non-blocking gate pushes the flip-to-blocking milestone further out.

## Ask

Chip the 85 down over one or more scoped PRs, lowering \`knip-baseline.json\` \`count\` in the same PR each time to lock in the gain (per the baseline's own guidance). Removing unused exports/types requires per-symbol analysis — some may be public API surface or referenced dynamically — so this is deliberately **not** a single bulk-delete.

Reproduce the current list:

\`\`\`
node_modules/.bin/knip --reporter json | jq '[.issues[] | {file, exports: (.exports // [] | map(.name)), types: (.types // [] | map(.name))}] | map(select((.exports|length)>0 or (.types|length)>0))'
\`\`\`

Once the count reaches 0, drop \`continue-on-error: true\` from the ratchet CI job to make it blocking.

Start by reading knip-baseline.json and running the provided knip command to inspect the 85 unused exports and types. Review each symbol's references before changing its owning source file, and update the baseline count with each scoped cleanup. Done means the count reaches 0 and continue-on-error: true is removed from the ratchet CI job.

Written by the indexing model from the issue text.