Preserve literal static attributes when folding

[joshhanley]

The Scenario

A user has a Flux radio group bound to a nullable boolean property using Livewire's .boolean modifier. Without Blaze, selecting the false radio option updates the property to false as expected. With Blaze enabled, selecting the false option still behaves as truthy, so the property appears to always be true.

<flux:radio.group wire:model.boolean.live="testBoolean" variant="cards">
    <flux:radio value="true" label="here is true" />
    <flux:radio value="false" label="here is false" />
</flux:radio.group>

The Problem

Blaze folding resolves static attribute values through Attribute::getStaticValue() before rendering folded output. That method was converting any static true, false, or null string into the equivalent PHP value, even when the attribute was not bound.

That meant value="false" became boolean false during folded rendering and was omitted by the attribute bag. Flux radio elements generate a fallback value when no value attribute is present, so the false option ended up with a generated non-empty string instead of "false".

Meanwhile, value="true" became boolean true and rendered as value="value". Livewire's .boolean modifier therefore received non-empty strings for both options, causing the false option to behave as truthy.

The Solution

This changes folded static attribute resolution so unbound literal values are kept as strings instead of being treated as PHP constants. Now value="false" stays "false" during folding and renders the same way it does in Blade.

Fixes livewire/flux#2651

[github-actions]

Benchmark Result: Default

Attempt	Blade	Blaze	Improvement
#1	338.90ms	15.66ms	95.4%
#2	352.78ms	15.84ms	95.5%
#3	349.62ms	15.70ms	95.5%
#4	348.13ms	15.96ms	95.4%
#5	338.83ms	15.67ms	95.4%
#6	343.66ms	15.49ms	95.5%
#7	346.79ms	16.01ms	95.4%
#8	344.46ms	15.82ms	95.4%
#9	350.18ms	15.78ms	95.5%
#10	345.21ms	15.64ms	95.5%
Snapshot	345.69ms	15.72ms	95.5%
Result	346.00ms (~)	15.74ms (~)	95.5% (~)

_{Median of 10 attempts, 5000 iterations x 10 rounds, 45.56s 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}