Phase 1: Make streaming pipeline chunk-emitting

[aram356]

Parent: #563

Goal

Make the internal streaming pipeline truly chunk-emitting so that processors emit output on every chunk instead of buffering the entire document. Ships independently with immediate memory savings.

Tasks

• Phase 1, Task 1: Fix encoder finalization in process_through_compression #568 — Task 1: Fix encoder finalization in process_through_compression
• Phase 1, Task 2: Convert process_gzip_to_gzip to chunk-based processing #569 — Task 2: Convert process_gzip_to_gzip to chunk-based processing
• Phase 1, Task 3: Convert decompress_and_process to chunk-based processing #570 — Task 3: Convert decompress_and_process to chunk-based processing
• Phase 1, Task 4: Rewrite HtmlRewriterAdapter for incremental streaming #571 — Task 4: Rewrite HtmlRewriterAdapter for incremental streaming
• Phase 1, Task 5: Full verification #572 — Task 5: Phase 1 full verification

Acceptance Criteria

• All compression paths (gzip, deflate, brotli) process chunks incrementally
• HtmlRewriterAdapter emits output per chunk, not only on is_last
• Encoder finalization errors propagate instead of being silently swallowed
• All existing tests pass, no regressions
• WASM build succeeds

Plan

See docs/superpowers/plans/2026-03-25-streaming-response.md (Tasks 1-5)

[aram356]

Replaced by individual task issues (#568-#572) under epic #563.