fix(binary): read the stale-series retry from the typed problem

[JosteinGj]

Replaces #131, which was opened six hours before the api changed under it.

What

is_stale_series_rejection searched the raw response body for unknown-timeseries or
external-id-mismatch, the idiom from before ResponseError::problem() existed. That also
matched the SDK's own pre-flight 404, which names the series it could not resolve, so an
external id spelling one of those tokens earned a pointless second attempt.

It reads the problem type now. Platform #120
(b3412ec0) gave each binary refusal the type its status calls for, so the slug discriminates
here the way it does on every other endpoint:

Slug	Status	Retried?
invalid-frame	400	no
unknown-timeseries	404	yes
request-too-large	413	no
unsupported-media-type	415	no
value-type-mismatch	422	no
external-id-mismatch	422	yes
too-many-in-flight	429	by post_frames, as before

The single datapoint-block-rejected that preceded them is deliberately not matched, and a
test says so: the binary path has never been in a release, so nothing can be sending it to a
caller of this SDK. reason survives server-side as the kebab-case sub-case and is what tells
the thirteen malformed-frame cases apart, but it is no longer the discriminator.

Why #131 could not just be merged

It gated on problem.slug() == Some("datapoint-block-rejected"), a slug the api no longer
emits, so merging it would have silently disabled the retry rather than fixed it. Its unit
test passed only because the fixtures hand-built the old body.

Verification

• only_a_stale_series_problem_rebuilds_the_request passes offline. Four cases, two of them
  negative on purpose: value-type-mismatch (re-resolving cannot make a value fit) and the
  retired datapoint-block-rejected, so a future compat fallback has to delete a test to land.
• test_insert_datapoints_binary_re_resolves_a_recreated_series is the first live coverage of
  the retry at all: every other path resolves before it sends. It is #[ignore]d next to
  test_datapoints_binary and needs a backend at platform test(python): POLICY type-labels are matchable now #120 or later — on an older one
  it fails with the pre-split body, which is the correct report, not a flake.
• CI here is cargo test --no-run, so green checks mean it compiles. They did on fix(binary): decide the stale-series retry from the problem document #131 too.

Notes for the reviewer

• Docs for the api-side split are already up as datahub-sdk-docs
  #80.
• The AGENTS.md hunk touches the same paragraph as docs/sync-with-sdk, so whichever lands
  second takes a small conflict. That branch also drops the stale "Not in the Python bindings
  yet" sentence, left alone here on purpose.

🤖 Generated with Claude Code