Benchmark transition to native PHP 8.4 HTMLDocument

[g105b]

The release of PHP 8.4 introduces much improved native DOM functionality. https://www.php.net/manual/en/class.dom-htmldocument.php

This should make a lot of GT's Dom redundant, and should increase speed due to it being native bindings.

However, it will be impossible to switch to the new native functionality without some work, and the work is imposed on GT along with other users of the DOM - all new native DOM classes are final. This means that GT DOM might have to take the decision to go back to proxying the native classes, rather than extending them, so that the GT classes can be used and extended by DomTemplate, and other projects.

---

Before any transition work can start, a benchmark needs to be defined, so we know what we're aiming for.

This should only be a few simple scripts that work with large documents or repetitive tasks on the smaller documents, but each script should run twice: once with the GT Dom, and once with the new native Dom. The actual code should be identical for both tests - in fact, there could be a dynamic instantiation of each class within the same script.

[g105b]

See benchmarks introduced in bc964a1 - over 90% speed increase, with no change required to the code.

[g105b]

Using the current implementation of Dom, here are my benchmark results:

01-basic
Gt\Dom\HTMLDocument: 0.00382
Dom\HTMLDocument: 0.00005
Speed increase: 99%

02-many-elements
Gt\Dom\HTMLDocument: 0.4553
Dom\HTMLDocument: 0.0081
Speed increase: 98%

03-nested
Gt\Dom\HTMLDocument: 0.6457
Dom\HTMLDocument: 0.0856
Speed increase: 87%

I have started experimenting with different implementations of wrapping the new dom. Here's my first benchmark results:

01-basic
Gt\Dom\HTMLDocument: 0.00038
Dom\HTMLDocument: 0.00002
Speed increase: 95%

02-many-elements
Gt\Dom\HTMLDocument: 0.0170
Dom\HTMLDocument: 0.0082
Speed increase: 52%

03-nested
Gt\Dom\HTMLDocument: 0.1679
Dom\HTMLDocument: 0.0794
Speed increase: 53%

So just with a few initial tweaks, the speed increase has reduced from ~100% to ~50%, meaning that the Gt\Dom library is now only half as slow as native, rather than 100% slower than native.

This is caused by there being a non-zero amount of setup time when using a wrapper class. The time spend in the GT code is probably not linear to the amount of work done, and is more likely to be a one off cost to working with the wrapped classes. But I don't think it's good enough, because I want to see the speed increases that native dom brings!