Parser optimizations

[asterite]

Hi!

See discussion starting at #536 (comment)

This is just a draft... I think the style or final code can be further improved, but at least the optimization ideas should be clear from the commits.

This is the benchmark I used:

require "./src/all"
require "benchmark"

filename = "icons.mint"
source = File.read(filename)

Benchmark.ips do |x|
  x.report("parse") do
    Mint::Parser.parse(source, filename)
  end
end

The file "icons.mint" comes from here: https://github.com/mint-lang/mint-tabler-icons/blob/as-functions/source/Icons.mint

Before this PR:

parse 108.63m (  9.21s ) (± 0.00%)  7.33GB/op  fastest

With this PR:

parse  68.05  ( 14.70ms) (± 3.80%)  20.1MB/op  fastest

[gdotdesign]

Thank you very much 🙏 I totally missed that raise, we have run into this before 🤦 I'll take a look at other optimizations you made to learn from it :)

[asterite]

Also with the previous benchmark:

require "./src/all"
require "benchmark"

filename = "./core/source/Provider/ElementSize.mint"
source = File.read(filename)
Benchmark.ips do |x|
  x.report("parse") do
    Mint::Parser.parse(source, filename)
  end
end

Before:

parse   8.55k (117.00µs) (± 4.06%)  83.9kB/op  fastest

After:

parse  15.92k ( 62.82µs) (± 6.46%)  37.2kB/op  fastest

So it's about twice as fast and consumes half the memory in the general case.

And for this particular icons.mint file, the speedup i like 600x faster and like 360 less memory consumed.

I still have to investigate why it takes 20MB to parse that file, so there's still room for optimization! But it can be done in a separate PR.

[Sija]

LGTM 🚀 (missing spec for next_whitespace_index would be a nice to have, but not required)

[gdotdesign]

Thank you very much! 🙏 🎉