Optimize {write,take}Leb128

[NicoElbers]

Old (Base impl) vs this (New impl)

Benchmarking methodology

Performance based on a micro-benchmark compiled with zig nightly build 0.16.0-dev.1234+74900e938 on ReleaseFast. Specifically zig build-exe bench.zig --name name -O ReleaseFast --zig-lib-dir lib.

The new impl was compiled with the standard library of da48ade while the old impl was compiled with the standard library of 416bf1d.

The 'small' category is values of at most 7 bits, the 'medium' category is values of at most 12 bits and the 'full' category is values of @bitSizeOf(T) bits.

For both datasets the benchmark was run after a system reboot and 2 full warmup runs.

Rewrite writeLeb128 to no longer use writeMultipleOf7Leb128, instead:

• Make use of byte aligned ints
• Special case small numbers (fitting inside 7 bits)

Rewrite Reader.takeLeb128 to not use takeMultipleOf7Leb128, instead:

• Use byte aligned integers
• Turn the main reading loop into an inlined loop of static length
• Special case small integers (<= 7 bits)

Writing performance:

• Amongst u8, u16, u32 and u64 performance gains are between ~1.5x and ~2x
• Amongst i8, i16, i32 and i64 performance gains are between ~2x and >4x
• For integers with bit multiples of 7 performance is roughly equal within the
  margin or error or slightly faster.

Reading performance:

• signed and unsigned 32 bit integers have 5x to 12x(!)
  performance improvement.
• For u8, u16 and u64 performance increases ~1.5x to ~6x
• For i8, i16 and i64 performance increases ~1.5x to ~3.5x
• For integers with bit multiples of 7 performance is roughly equal within the
  margin or error or slightly faster.