fix(ai-speech2txt): charge Gemini transcriptions that fail after Google answers

Google bills an interaction it processed whether or not it completed, so an
incomplete, failed or malformed reply was free to the caller and paid by
Puter; audio whose transcript overran the output cap could be resubmitted at
no cost. Usage is now recorded before the reply is checked, falling back to
the counted audio tokens when the reported usage can't be read.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/docs/src/AI/speech2txt.md

@@ -64,7 +64,7 @@ Gemini rejects options it can't honor with `bad_request` instead of ignoring the
 
 Gemini accepts WAV, MP3, AIFF, AAC, OGG, FLAC, MPEG, M4A, L16, Opus, A-law, mu-law and WebM audio, sent with an `audio/*` MIME type (`audio/wav`, `audio/mpeg`, `audio/webm`, and so on). Other types fail with `bad_request`, and the audio isn't converted. The audio can be at most 10 MiB, or 1 hour long, or 30 minutes long with word timestamps or `diarize`. Gemini enforces the duration limits itself, and the call fails when the audio is longer. A transcript is limited to 8,192 output tokens. A longer one fails with `502 upstream_failed`, so split long recordings into shorter files.
 
-Before transcribing, Puter reserves the cost of the counted audio tokens plus the full 8,192 output tokens, then charges the tokens Gemini reports.
+Before transcribing, Puter reserves the cost of the counted audio tokens plus the full 8,192 output tokens, then charges the tokens Gemini reports. A transcription Gemini answers is charged even when it fails, as Gemini bills it, including one cut off at the output limit. When Gemini reports no usable token counts, only the counted audio is charged.
 
 #### `testMode` (Boolean) (optional)