Fixes #385 — pages served with gzip/brotli/zstd content-encoding (e.g. uindex.org, 1337x.to) returned garbled Unicode in solution.response after the v3.0.0 CSP-stripping change.
Root cause: the CSP-strip route rewrites document responses via route.fetch() + route.fulfill(). route.fulfill re-serves the raw compressed bytes fetched over the wire while the forwarded headers no longer match — the browser renders compressed bytes as text.
Fix: fetch document bodies with accept-encoding: identity so the re-served body is plain text, and strip stale content-encoding/content-length headers alongside CSP.
Verified in CI (docker build --target test): 23 passed, 0 failed — including the real-site test_bypass suite.