Add fast paths for class/style string arguments in BlazeAttributeBag

[SanderMuller]

When merge(), class(), or style() receive a simple string argument (the most common usage), skip the general-purpose loops and handle the merge directly.

The shared class-merging logic is extracted into withMergedClass() to avoid duplication between merge() and class().

This replaces both PR #157 and #161 with a single combined PR.

[SanderMuller]

/benchmark class

[SanderMuller]

/benchmark merge

[github-actions]

Benchmark Result: Class

Attempt	Blade	Blaze	Improvement
#1	240.77ms	13.68ms	94.3%
#2	234.65ms	13.61ms	94.2%
#3	236.62ms	13.60ms	94.3%
#4	228.92ms	13.25ms	94.2%
#5	237.33ms	13.55ms	94.3%
#6	231.15ms	13.47ms	94.2%
#7	234.94ms	13.41ms	94.3%
#8	230.39ms	13.66ms	94.1%
#9	231.89ms	13.48ms	94.2%
#10	235.80ms	13.38ms	94.3%
Snapshot	235.27ms	17.91ms	92.4%
Result	234.80ms (~)	13.52ms (-24.5%)	94.2% (+1.8%)

_{Median of 10 attempts, 5000 iterations x 10 rounds, 31.77s total}

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

[github-actions]

Benchmark Result: Merge

Attempt	Blade	Blaze	Improvement
#1	170.01ms	11.64ms	93.2%
#2	166.45ms	11.74ms	92.9%
#3	165.58ms	11.71ms	92.9%
#4	169.72ms	11.69ms	93.1%
#5	169.87ms	11.62ms	93.2%
#6	167.13ms	11.70ms	93%
#7	169.59ms	11.65ms	93.1%
#8	169.31ms	11.62ms	93.1%
#9	166.73ms	11.73ms	93%
#10	166.29ms	11.73ms	92.9%
Snapshot	166.13ms	13.64ms	91.8%
Result	168.22ms (~)	11.70ms (-14.2%)	93% (+1.2%)

_{Median of 10 attempts, 5000 iterations x 10 rounds, 23.32s total}

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

[github-actions]

Benchmark Result: Default

Attempt	Blade	Blaze	Improvement
#1	339.93ms	15.92ms	95.3%
#2	340.90ms	16.09ms	95.3%
#3	350.71ms	15.87ms	95.5%
#4	351.00ms	15.89ms	95.5%
#5	347.52ms	15.98ms	95.4%
#6 *	354.52ms	16.57ms	95.3%
#7	343.42ms	15.97ms	95.3%
#8	341.95ms	15.77ms	95.4%
#9	347.23ms	15.72ms	95.5%
#10	343.08ms	16.12ms	95.3%
Snapshot	345.08ms	15.91ms	95.4%
Result	343.42ms (~)	15.92ms (~)	95.4% (~)

_{Median of 10 attempts (* = outlier, excluded from result), 5000 iterations x 10 rounds, 45.63s total}

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

[ganyicz]

Looks like this causes an attribute ordering issue in the FluxPro tests (you would need to dev-require Flux Pro with a valid license to run these):

Tests\FluxProTest > chart:

- <ui-chart wire:model="data" class="block [:where(&amp;)]:relative w-full aspect-2/1" wire:ignore.children >
+ <ui-chart class="block [:where(&amp;)]:relative w-full aspect-2/1" wire:model="data" wire:ignore.children >

Alternatively feel free to add a separate comparison test for this

[SanderMuller]

Looks like this causes an attribute ordering issue in the FluxPro tests (you would need to dev-require Flux Pro with a valid license to run these):

Tests\FluxProTest > chart:

- <ui-chart wire:model="data" class="block [:where(&amp;)]:relative w-full aspect-2/1" wire:ignore.children >
+ <ui-chart class="block [:where(&amp;)]:relative w-full aspect-2/1" wire:model="data" wire:ignore.children >

Alternatively feel free to add a separate comparison test for this

Thanks. I do have a Flux Pro license, but figured it would be good to have a CI test so I took the option to add a test for it (and make the code pass it)