perf: special case small integers in fmt

[DaniPopes]

This is a very cheap check to avoid using to_base_be at all when the integer is small enough. It doesn't show any improvements in the benchmarks for >128 bits because the numbers are uniformly sampled, so they will always be at full bits, but it's a massive speed up as can be seen from the 64 and 128 bit benchmarks.

Based on #490.

[codspeed-hq]

CodSpeed Performance Report

Merging #500 will improve performances by ×2.3

_{Comparing DaniPopes:perf-fmt-minor (102e883) with main (2973b98)}

Summary

⚡ 7 improvements
✅ 288 untouched benchmarks

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	fmt/binary/64	4.2 ms	2.6 ms	+59.02%
⚡	fmt/decimal/128	4.8 ms	2.3 ms	×2.1
⚡	fmt/decimal/64	2.8 ms	1.3 ms	×2.1
⚡	fmt/hex/128	5.3 ms	3 ms	+76.58%
⚡	fmt/hex/64	3.5 ms	1.5 ms	×2.3
⚡	fmt/octal/128	4.9 ms	3.1 ms	+59.58%
⚡	fmt/octal/64	2.9 ms	1.5 ms	+99.5%