perf: use bit shifts for power-of-two formatting

[DaniPopes]

No description provided.

[codspeed-hq]

Merging this PR will improve performance by ×7.7

⚡ 12 improved benchmarks
✅ 374 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	fmt/binary/4096	5.3 ms	2.3 ms	×2.3
⚡	fmt/hex/384	11.2 ms	6.3 ms	+76.96%
⚡	fmt/hex/256	8 ms	4.5 ms	+77.12%
⚡	fmt/hex/192	6.4 ms	3.6 ms	+80.5%
⚡	fmt/hex/4096	4,519.7 µs	584.9 µs	×7.7
⚡	fmt/octal/512	15.8 ms	13.7 ms	+15.13%
⚡	fmt/octal/256	8.6 ms	7.3 ms	+18.66%
⚡	fmt/hex/512	14.4 ms	8.1 ms	+77.43%
⚡	fmt/octal/4096	4.4 ms	1.3 ms	×3.5
⚡	most_significant_bits/4096/4096	28.9 µs	26.1 µs	+10.88%
⚡	to/f64/4096	29.4 µs	26.3 µs	+11.4%
⚡	to/f32/4096	29.2 µs	26.3 µs	+11.1%

---

_{Comparing DaniPopes:dani/fmt-pow2-fast-paths (edd7b13) with main (c147f59)}