chore: prepare 0.1.0 release (LICENSE, changelog, release workflow, Homebrew docs)

[TakiTake]

Summary

Prepares pall8t for its first public release (v0.1.0):

• Version reset: Cargo.toml/Cargo.lock from 0.4.0 to 0.1.0. No tags or releases were ever published, so this reset is safe.
• LICENSE: standard MIT text, copyright TakiTake 2026 (matches the license = "MIT" already declared in Cargo.toml).
• CHANGELOG.md: Keep a Changelog format, one ## [0.1.0] - 2026-07-12 entry summarizing the current feature set (sandboxed agent runs, persistent shared home, content-hash rebuilds + pruning, worktree awareness, reference-repo duplication, herdr integration, home compositor isolated mode, the full CLI surface).
• .github/workflows/release.yml: triggers on v* tag push, verifies the tag matches Cargo.toml's version (via cargo pkgid), builds natively on macos-latest (arm64), sanity-checks pall8t --version output, packages a pall8t-vX.Y.Z-aarch64-apple-darwin.tar.gz (binary + LICENSE + README) with a .sha256 checksum, extracts release notes from the matching CHANGELOG section, and publishes via gh release create.
• README.md: leads the install instructions with brew install TakiTake/tap/pall8t (annotated as shipping with the v0.1.0 release), keeps cargo install --path . as the from-source alternative, and trims "Rust toolchain" from the hard requirements list.
• docs/release.md: maintainer-facing release process doc, including a note that the Homebrew tap repo itself needs to be created (not just updated) on the first release.

Notes

• The Homebrew tap repo (TakiTake/homebrew-tap) doesn't exist yet — creating it and its formula is a separate, later task (tracked as task Server/client architecture: central state daemon + detach/reattach #2), not part of this PR.
• No tags are pushed by this PR; tagging v0.1.0 happens after this merges.

Quality gates (all passing in the worktree)

• cargo fmt --check
• cargo clippy --all-targets -- -D warnings (host)
• cargo clippy --all-targets --target aarch64-apple-darwin -- -D warnings (cross-target, since the repo has #[cfg(target_os)] branches)
• cargo test — 134 + 1 tests pass
• .github/workflows/release.yml YAML parses cleanly (validated with js-yaml); the tag/version-match and CHANGELOG-extraction shell logic was dry-run locally against this repo's real Cargo.toml/CHANGELOG.md
• Reviewed with /code-review and /skeptical-review — the skeptical review caught a real README defect (the install quickstart's working cargo install --path . command was commented out, so copy-pasting the block installed nothing) which is fixed in this PR; /simplify applied three follow-on improvements to the release workflow (fail-fast CHANGELOG check before the build, dependency caching via Swatinem/rust-cache, and cargo pkgid instead of hand-parsing Cargo.toml for the version)

Test plan

• After merge, task Server/client architecture: central state daemon + detach/reattach #2 tags v0.1.0, confirms the release workflow runs end-to-end on a real tag push, and creates TakiTake/homebrew-tap with a formula pointing at the published binary + checksum

🤖 Generated with Claude Code