feat: readable rb ffprobe output

[a-essawy]

Summary

rb ffprobe currently prints a single terse line on success (e.g. Video mp4 0:05 1280x720 24fps), dropping almost every useful field the probe actually returns. This replaces it with a clean, multi-line summary block.

• New formatSummaryBlock() renders a header line (kind, container, duration, size) plus a labeled row per stream kind (Video/Audio/Image), adapting to what the media actually has:
  • video: codec/profile, resolution + aspect ratio, fps (flags VFR), bit depth + pixel format, bitrate, HDR flag, rotation, language
  • audio: codec/profile, channel layout (or raw count), sample rate, bitrate, language
  • image: dimensions, bit depth + pixel format
  • other: stream counts (skipping zero counts)
  • only present fields are rendered; nothing prints as undefined
• New resolveOutputText() handles the two shapes output.data can take: a parsed summary renders the block above; a summary-less response (raw -print_format csv/xml/-of default) prints output.data.stdout verbatim instead, since that's exactly what the user asked ffprobe for.
• --json still prints the full output.data, and the trailing dashboard job URL line is unchanged.
• Expanded the ProbeSummary/ProbeData local types to match the fuller documented output.data contract (profile, displayAspectRatio, bitDepth, channelLayout, sampleRate, isHdr, rotation, language, etc.) instead of the minimal subset the old one-liner used.

Test plan

• pnpm typecheck clean
• pnpm test — 174 pass, 0 fail
• New tests cover: video summary (codec+resolution+fps present), audio-only summary (audio row present, no Video row), raw-stdout response (prints stdout verbatim, no summary block rendered)