Refactor to AST

[ganyicz]

Following #193, which adds support for parsing directives, this PR refactors rest of Blaze to take advantage of that:

• Compiling @props and @aware directives
• Compiling @unblaze blocks
• Analyzing @blaze, @props and @aware directives
• Extracting use statements from @php blocks and @use directives

Previously we used DirectiveCompiler for this. This class hacked into Laravel's BladeCompiler to delegate the parsing of directives to Blade, however in #193 we brought most of that logic into our tokenizer and directives are now part of the AST.

After this PR, all analysis and compilation is now driven by the AST, including skipping over <?php, @php and @verbatim blocks, which was previously also done by hacking into BladeCompiler.

This required bigger structural changes however creates multiple benefits:

• Faster compilation: making Blaze only ~5% slower than Blade (prev. 900% slower)
• Unified approach: everything is now driven by the AST
• Dropping reliance on internal state of BladeCompiler in most places

The performance improvement is mostly the result of caching the AST and dropping the DirectiveCompiler which used reflection and had to run multiple regexes over the entire template repeatedly.

[github-actions]

Benchmark Result: Default

Attempt	Blade	Blaze	Change
#1	365.34ms	14.73ms	96%
#2	363.56ms	15.02ms	95.9%
#3	362.06ms	14.69ms	95.9%
#4	364.47ms	14.87ms	95.9%
#5	368.33ms	14.83ms	96%
#6	361.97ms	15.31ms	95.8%
#7	360.68ms	15.04ms	95.8%
#8	366.10ms	14.66ms	96%
#9	366.85ms	14.72ms	96%
#10	367.64ms	14.89ms	95.9%
Snapshot	363.13ms	14.70ms	96%
Result	364.91ms (~)	14.85ms (~)	95.9% (~)

_{Median of 10 attempts, 5000 iterations x 10 rounds, 47.83s total}

_{To run a specific benchmark, comment /benchmark <name>
attributes, aware, class, default, forwarding, merge, named-slots, no-attributes, slot, compilation}

[ganyicz]

/benchmark compilation

[github-actions]

Benchmark Result: Compilation

Attempt	Blade	Blaze	Change
#1	2.21ms	2.33ms	-5.4%
#2	2.27ms	2.35ms	-3.5%
#3	2.22ms	2.34ms	-5.4%
#4	2.24ms	2.34ms	-4.5%
#5	2.23ms	2.34ms	-4.9%
#6	2.24ms	2.36ms	-5.4%
#7	2.22ms	2.32ms	-4.5%
#8	2.24ms	2.36ms	-5.4%
#9	2.27ms	2.37ms	-4.4%
#10	2.25ms	2.36ms	-4.9%
Snapshot	2.20ms	21.11ms	-859.5%
Result	2.24ms (~)	2.34ms (-88.9%)	-4.5% (+855%)

_{Median of 10 attempts, 5000 iterations x 10 rounds, 2.87s total}

_{To run a specific benchmark, comment /benchmark <name>
attributes, aware, class, default, forwarding, merge, named-slots, no-attributes, slot, compilation}

[ganyicz]

/benchmark compilation

[ganyicz]

/benchmark compilation

[ganyicz]

/benchmark compilation

[ganyicz]

/benchmark compilation