Add colored blockquotes! (Markdown preprocessor) + Benchmarks

[caendesilva]

See #483 (comment)

[caendesilva]

The system I'm working on now is pretty cool, but adds a lot of complexity. We'll see how it goes, maybe it belongs in a dedicated package. (crosspost from #483 (comment))

[caendesilva]

[caendesilva]

Benchmarks for the feature as it is now:

Note this is a rather crude benchmark, rand() is used to try to spoof PHP from caching the results.

System info for all tests

PHP version	OS/Arch	xdebug	opcache
8.0.13 (cli)	WINNT (64-bit)	enabled ✅	disabled ❌

The test cases

1 Benchmark baseline for parsing normal blockquote

Benchmark::run(function () {
    echo (new \Hyde\Framework\Services\MarkdownConverterService('> '.rand(0, 1000)))->parse();
}, 10000, 'Baseline Test');

2 Benchmark baseline for parsing normal blockquote with the entire preprocessing system disabled

Benchmark::run(function () {
    echo (new \Hyde\Framework\Services\MarkdownConverterService('> '.rand(0, 1000)))->parse();
}, 10000, 'Baseline No-Processing');

3 Benchmark for the colored blockquote preprocessor

Benchmark::run(function () {
    echo (new \Hyde\Framework\Services\MarkdownConverterService('>info '.rand(0, 1000)))->parse();
}, 10000, 'Colored Blockquotes');

Results (single line tests)

Name	Total iterations	Total execution time	Avg. iteration time	Avg. iterations/sec	Approx. Memory usage
1 Baseline Test	10 000	29847.73ms	2.98477252ms	0.34	14MB
2 Baseline No-Processing	10 000	26405.78ms	2.64057779ms	0.38	14MB
3 Colored Blockquotes	10 000	27512.21ms	2.75122061ms	0.36	14MB

Result analysis

It's expected that the second case without any processing is the fastest. What I did not expect is that the colored blockquotes are faster. I wonder if this is because my test somehow is misrepresenting, though it in a way makes sense as it is very possible that my Markdown parser is faster than the CommonMark one. Either way, in this test scenario the difference is negligible, though I want to test with a multiline dataset.