Fix VP8 encoder mode/coefficient ordering to match decoder expectations

[Copilot]

VP8 decoder reads all macroblock modes via vp8_decode_mode_mvs() before reading any coefficients via decode_mb_rows(). The encoder was interleaving mode and coefficient writes per macroblock, causing decoder rejection.

Changes

Restructured keyframe encoding into two phases:

• Phase 1: Write Y-mode and UV-mode for all macroblocks using vp8_treed_write() with proper probability trees
• Phase 2: Write coefficient data (16 Y + 4 U + 4 V blocks) for all macroblocks

Created vp8e_encode_macroblock_coeffs() to separate coefficient encoding from mode encoding logic.

// Before: interleaved per macroblock
for (mb_row...) {
  for (mb_col...) {
    encode_mode_and_coeffs(mb);  // Wrong order
  }
}

// After: two-phase encoding
for (mb_row...) {
  for (mb_col...) {
    write_ymode(); write_uvmode();  // Phase 1: all modes
  }
}
for (mb_row...) {
  for (mb_col...) {
    write_all_coeffs();  // Phase 2: all coefficients
  }
}

Frame structure verified: partition size correctly calculated, boolean encoder writing to proper buffer offset, tree-based mode encoding producing correct bit sequences.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[paulomorgado]

Do not use byte[] or byte*. Use Span<byte> or ReadOnlySpan<byte> instead.