Migrate to released Mojo 1.0 and cut 0.2.0

[conorbronsdon]

Mojo 1.0 shipped, so this comes off nightly. pixi moves to the max channel with mojo >=1.0,<2, and CI installs mojo==1.0.0 from PyPI instead of the nightly index with --prerelease allow. The weekly drift check still points at the latest nightly, so the repo keeps telling us when the language moves under it.

One source fix, benchmark only: perf_counter_ns() no longer returns UInt, so bench/bench_resp.mojo failed to compile until _report took an Int.

The README's test counts were stale: 34 → 36 protocol tests, and the two transport regressions CI also runs are now named.

Cuts 0.2.0.

Verified locally on Mojo 1.0.0: 36 protocol + 2 transport tests green, 21 integration tests green against a live redis-server, mojo format clean, benchmark builds and runs.

[conorbronsdon]

Verification notes (generated — detail for reviewers and review tooling)

All of the below ran locally on Mojo 1.0.0 (ed45d567), Ubuntu on WSL2, against this branch.

• Tests: test_resp 36 and test_connection 2 (both run in CI), plus test_integration 21 against redis-server --port 6399 — 0 failed anywhere.
• The benchmark break is a real 1.0 regression, not a stale pin: bench/bench_resp.mojo builds and runs under the b3 nightly the repo was pinned to and fails to compile under 1.0.0 with cannot be converted from 'Int' to 'UInt', because perf_counter_ns() changed return type.