[nexxtmobile]

Fixes #12747

The client-commit path (commitUtteranceWithTranscript) and writeUtteranceWAV
wrote the utterance WAV with a fixed 16 kHz header (laudio.NewWAVHeader),
while the input buffer holds audio at the session input sample rate (e.g.
24 kHz for OpenAI-Realtime clients). The file-based transcription backend
decoded 24 kHz data at 16 kHz (1.5x speed/pitch) → garbage/empty transcripts
→ input_not_committed → total silence after the greeting.

writeUtteranceWAV now takes the sample rate; both call sites pass
session.InputSampleRate (via the existing laudio.NewWAVHeaderWithRate).

Includes a regression spec: the committed WAV header carries the requested
rate (24 kHz and 16 kHz).

Tested: go test -race ./core/http/endpoints/openai/ green (Go 1.26).

[localai-org-maint-bot]

@mudler Code-review sign-off at 7621916df, subject to CI and human DCO certification of the documentation fix. The committed-utterance WAV now uses the session input rate on both file-based transcription paths; all helper callers are updated. I added the required user-facing documentation. Reviewed against master 48f358dd9; whitespace checks pass. Focused endpoint tests were blocked by a dependency-proxy HTTP Forbidden response for LocalAGI; race testing also lacks CGO/a C compiler. No claim that runtime tests passed.

[nexxtmobile]

Update: rebased onto mudler:master e51747055 (the branch had gone stale / conflicting). go test -race ./core/http/endpoints/openai/ is green on the new base.

Note on the doc commit: a rebase force-push clobbered the maint-bot's doc commit (7621916df). I have re-applied the identical doc change (same file, same text) as 5859326bf with human DCO attestation, per the process used in #12622. Sorry for the churn.