Faster integer formatting

[radiospiel]

This PR provides an alternative implementation for a long → decimal conversion.

The main difference is that it uses an algorithm pulled from https://github.com/jeaiii/itoa. The source there is C++, it was converted by hand to C for inclusion with this gem. A writeup of this algorithm can be found here. jeaiii's algorithm is covered by the MIT License, see source code.

On addition this version now also generates the string directly into the fbuffer, foregoing the need to run a separate memory copy.

As a result, I see a speedup of 32% on Apple Silicon M1 for an integer set of benchmarks.

== Encoding ints (45025 bytes)
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +YJIT +PRISM [arm64-darwin24]
Warming up --------------------------------------
        json (local)     1.566k i/100ms
Calculating -------------------------------------
        json (local)     16.322k (± 3.9%) i/s   (61.27 μs/i) -     82.998k in   5.093499s

Normalize to 8009 byte
== Encoding ints (45025 bytes)
ruby 3.4.1 (2024-12-25 revision 48d4efcb85) +YJIT +PRISM [arm64-darwin24]
Warming up --------------------------------------
       json (2.10.2)     1.228k i/100ms
Calculating -------------------------------------
       json (2.10.2)     12.352k (± 0.5%) i/s   (80.96 μs/i) -     62.628k in   5.070472s

These benchmarks are run via a script (link) which is based on the gem's benchmark/encoder.rb file. There are probably better ways to run benchmarks :) My version allows to combine multiple test cases into a single one.

The script adds a "cents" test case (link) which explicitly covers integers in a range that is relevant for financial institutions. (Comparing benchmarks for that case show a speedup of 41%)

The dumps benchmark, which covers the JSON files in benchmark/data/*.json, sees a speedup of ~8 %.

[radiospiel]

Note: see this for discussing licensing concerns. Marking this PR as draft for the time being.

[radiospiel]

The windows build errors have been fixed. This PR is ready for review.

The changes are most impactful for long integers. A tests on 19 digit numbers shows a 65% speedup; the "cents" test (a mix of numbers between 1 and 7 digits) has 41%. On 2-digit numbers the gains are still ~11%. (all tested on macOS, Apple Silicon M1)

I think a 32-bit version could be built and might show some small gains. The C++ version, which is the base for these changes, specialises templates for various integer widths. I didn't want to go /that/ far here.