docs: remove internal provider and stack disclosures

[a-essawy]

Public-surface leak-audit remediation for the docs. These pages (and the Mintlify-generated /docs/llms-full.txt) named the internal provider stack and the cost-plus margin model, which a competitor can use to reproduce the service.

Changes

• job-types/{ffmpeg,caption-burn,captions-animate}.mdx: remove the **Provider:** triggerdev field from the metadata line (users cannot pick the provider; it only named the orchestration vendor).
• concepts/job-lifecycle.mdx: drop the (Trigger.dev) parenthetical from the dispatched state.
• job-types/captions-animate.mdx: genericize the transcription pipeline prose (was "transcribed by whisper.cpp ... rendered as ASS subtitles") and the whisper captions keyword. Keeps the user-facing outcome (auto-transcribed, word-level, burned-in).
• concepts/credits.mdx + support/faq.mdx: remove the cost-plus formula ("actual provider cost plus infrastructure cost plus margin"). Billing language now states jobs are charged for the compute consumed.
• guides/ffmpeg.mdx: replace the enumerated 8-layer sandbox blocklist (exact blocked formats/filters/patterns + the -protocol_whitelist detail + the sandbox vendor name) with a 4-point trust statement. The enumeration doubled as an attacker bypass checklist.

Verification

• Source grep for provider/stack tells (triggerdev, trigger.dev, whisper, deepgram, videoseal, openalpr, mediapipe, cost-plus): none remain
• FAQ FAQPage JSON-LD still parses (9 questions, round-trips)
• Job-type metadata lines have no orphaned · separators

Note: mint validate / mint broken-links were not run locally (CLI not installed); Mintlify validates on deploy.