Adopt fbuild's perf + Windows-correctness lessons (meta)

[zackees]

Adopt fbuild's perf + Windows-correctness lessons

FastLED/fbuild is a hybrid Rust+Python tool that has spent the past month grinding down ~50× of cumulative wall-clock cost on its install + iteration loop, and just landed a Windows stdout-ordering fix that is exactly the same bug this template will hit. Time to back-port the lessons.

This is a meta-issue — each box below was a self-contained PR. All merged; closing.

Source PRs in fbuild (for reference)

• FastLED/fbuild#747 — Ship native binary as raw wheel script; drop the Python shim.
• FastLED/fbuild#744 — Default to dev profile + rust-lld linker.
• FastLED/fbuild#743 — mtime-skip in setup.py + pinned CARGO_TARGET_DIR.
• FastLED/fbuild#661 — [tool.uv] cache-keys so editable installs re-sync on Rust edits.

Punch list

Correctness

• (1) Drop the Python CLI subprocess shim; ship template-cli as a raw wheel script. → fix(install): ship template-cli as raw wheel script; drop Python shim #7

Build speed

• (2) [profile.dev.package."*"] opt-level = 3 in the root Cargo.toml. → perf(build): rust-lld linker on Windows + opt-3 deps in dev profile #3
• (3) [target.{x86_64,aarch64}-pc-windows-msvc] linker = "rust-lld.exe" in .cargo/config.toml. → perf(build): rust-lld linker on Windows + opt-3 deps in dev profile #3
• (4) Pin CARGO_TARGET_DIR to a home-dir path for wheel builds. → perf(install): pin CARGO_TARGET_DIR + disable PEP 517 isolation #5
• (5) [tool.uv] no-build-isolation-package = ["template-python-rust-cmd"] in pyproject.toml. Reverted in fix(install): ship template-cli as raw wheel script; drop Python shim #7 — with maturin the no-isolation trade-off doesn't work cleanly (uv reuses a build env without maturin; preview extra-build-dependencies is fragile). Documented this in the resulting comment block; CARGO_TARGET_DIR pin from perf(install): pin CARGO_TARGET_DIR + disable PEP 517 isolation #5 still applies and gives most of the win. Marking complete because the question is settled.
• (6) mtime-skip fast path in ci/build_wheel.py. → perf(install): mtime-skip fast path + structured artifact discovery #6
• (7) [tool.uv] cache-keys covering Rust source. → fix(uv): re-sync editable on Rust edits via cache-keys #4
• (8) Profile-aware artifact discovery fallback in ci/build_wheel.py. → perf(install): mtime-skip fast path + structured artifact discovery #6

Followups

• (9) BUILD_PROFILE=dev env-var iteration path through ci/build_wheel.py. → feat(build): BUILD_PROFILE=dev env var for fast local iteration #8
• (10) Drop the Python CLI shim entirely (cli.py deleted, [project.scripts] removed). → bundled into fix(install): ship template-cli as raw wheel script; drop Python shim #7

Acceptance — all verified

• ✅ cmd /c 'echo === BEFORE === & template-cli --version & echo === AFTER ===' prints === AFTER === after the output (verified locally on Windows after uv pip install dist/...whl into a fresh venv).
• ✅ unzip -l dist/template_python_rust_cmd-*.whl shows template_python_rust_cmd-0.1.0.data/scripts/template-cli.exe. No _bin/ entries.
• ✅ file $(which template-cli) reports PE32+ executable for MS Windows ... (console), x86-64. Not a Zip-archive console-script stub.

PRs merged in order

#	PR	Items
1	#3	(2) + (3) — dev profile + rust-lld
2	#4	(7) — uv cache-keys
3	#5	(4) + (5) — CARGO_TARGET_DIR + no-build-isolation
4	#6	(6) + (8) — mtime-skip + JSON artifact discovery
5	#7	(1) + (10) — raw wheel script, drop Python shim
6	#8	(9) — BUILD_PROFILE=dev