docs: Fix documentation mismatches with actual client behavior

[vdusek]

Fixes 12 documentation issues found by an engineering audit, where docs, examples, or docstrings contradict what the client actually does:

• The custom HTTP client guide example (HttpxClient) now raises ApifyApiError for error responses instead of returning them raw, and the guide documents this part of the call contract (resource clients rely on it, e.g. to translate a 404 into a None return value).
• The streaming concepts page no longer claims all three streaming methods yield a raw impit.Response — it now describes the actual yielded value per method (stream_record yields a dict with the response under value, stream and stream_record may yield None).
• The pagination concepts page no longer lists ListOfRequests among page models exposing total/offset/count; it now explains its cursor-based pagination via next_cursor.
• The logging formatter example no longer references %(status_code)s (absent on most records, causing logging errors) and no longer attaches a duplicate handler; the page notes which properties are present on every record.
• The upgrading-to-v3 guide cross-links now include the site baseUrl (/api/client/python/...), fixing 6 links that 404ed on the published site.
• The conda instruction for the brotli extra installs brotli-python (the Python bindings) instead of brotli (the C library).
• RunClient.resurrect docstrings cite the real SUCCEEDED status instead of the nonexistent FINISHED.
• wait_for_finish docstrings in RunClient and BuildClient spell the terminal status as TIMED-OUT (the real literal) instead of TIMED_OUT.
• The quick-start page refers to the Run model's default_dataset_id attribute instead of the v2-era run dictionary with defaultDatasetId.
• The README dataset example passes fields as list[str] per the signature instead of a comma-separated string.
• The README quick-start examples handle the Run | None return of call() instead of accessing attributes on a possible None.
• The timeouts concepts page states that no_timeout is capped at 24 hours by the default client instead of claiming it disables the timeout entirely.

✍️ Drafted by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.64%. Comparing base (2adf515) to head (c15a2d1).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #958   +/-   ##
=======================================
  Coverage   94.64%   94.64%           
=======================================
  Files          58       58           
  Lines        5263     5263           
=======================================
  Hits         4981     4981           
  Misses        282      282           

Flag	Coverage Δ
integration	92.30% <ø> (ø)
unit	83.94% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Pijukatel]

Conda forge does not have any notion of extras dependencies. So I am not sure this will work. It would require some testing. Maybe better to drop all brotli-related advice until tested.

I created a follow-up:
conda-forge/apify-client-feedstock#3

[vdusek]

Yes, it seems Conda-forge doesn't recognize the extras, so installing Brotli directly is the equivalent, isn't it?

Maybe I can put an additional note there to clarify that.