fix: honor BASE_TIMEOUT/BASE_MAX_RETRIES in scrape_as_html and extract

[karaposu]

Problem

BASE_TIMEOUT and BASE_MAX_RETRIES are documented, operator-facing settings applied by the
base_request wrapper — but scrape_as_html and extract's scrape stage POST to the same unlocker
/request endpoint with raw axios, so the two knobs are silently ignored there. An operator who
sets BASE_TIMEOUT=30 sees it "work sometimes": markdown scrapes time out properly while HTML scrapes
and extract hang indefinitely, with nothing marking the difference.

Fix

Route both calls through base_request — a one-token swap at each site; both configs are already
field-identical to scrape_as_markdown's wrapped call. With the env vars unset, behaviour is
byte-identical (one attempt, timeout: 0 = axios's own default, same error instances rethrown). With
them set, both tools now time out and retry exactly as their siblings do on the same endpoint.

Behaviour change (only when the knobs are configured)

Worst case per scrape becomes (retries+1) × timeout, and a retried scrape re-issues the request —
matching the existing contract of the four already-wrapped tools. A timeout surfaces as a clear error
instead of an indefinite hang.

Tests

test/base-request-coverage.test.js: a source-level invariant that every
api.brightdata.com/request call site uses base_request. It fails on the pre-fix code (2 of 6 sites
raw) and guards against a future tool being added with a raw axios /request call.

Notes

• Best merged after Fix six runtime defects: free-tier crash, NaN config, duplicate browser refs, erased errors, failed-job-as-data, stale-token attribution #163, which fixes the BASE_MAX_RETRIES parsing this wrapper depends on
  (on current main, a malformed value NaN-poisons the retry loop for all wrapped tools); the fixes
  are otherwise independent.
• fix: resolve extract's sampling session from the calling context instead of server.sessions[0] #168 edits the sampling half of the same extract function — whichever merges second is a trivial
  rebase.