Motivation

• Make fuzz orchestration more ergonomic for local and PR workflows by adding list and all modes and a sane default target.
• Prevent silent drift between fuzz/fuzz_targets/*.rs and fuzz/Cargo.toml so newly added harnesses are not omitted from CI or PR evidence.
• Avoid expensive key-generation fuzz runs in routine PR evidence while still compiling all harnesses to catch regressions early.

Description

• Add CLI switches to xtask fuzz: --list to print manifest-declared fuzz targets and --all to run every declared target, and keep rsa_pkcs8_pem_parse as the default when no target is given (DEFAULT_FUZZ_TARGET).
• Implement manifest-backed discovery and validation in list_fuzz_targets() by parsing fuzz/Cargo.toml, verifying each [[bin]] entry, and rejecting undeclared fuzz_targets/*.rs sources.
• Centralize cargo-fuzz availability check in ensure_cargo_fuzz_available() so --list can run without cargo-fuzz installed while real fuzz runs still fail fast if missing.
• Update PR fuzz step (fuzz_pr) to cargo fuzz build all harnesses for compile coverage then run a bounded smoke subset via is_pr_fuzz_target() to skip expensive keygen or chain-building harnesses.
• Add helper is_pr_fuzz_target() to exclude expensive targets and extend xtask unit tests to cover manifest-backed discovery, manifest-drift detection, PR-target filtering, and list behavior.

Testing

• Ran cargo test -p xtask fuzz -- --nocapture and cargo test -p xtask fuzz_targets -- --nocapture and all xtask unit tests related to fuzz discovery and filtering passed.
• Ran cargo run -p xtask -- fuzz --list to exercise the new listing path and it printed the declared fuzz targets (manifest-backed); this succeeded.
• Ran cargo check -p xtask to verify compilation and static checks; this succeeded.

Codex Task