Improvement: Add Cache layer to Graph::build steps

[chrispenny]

Fixes #14
Closes #14

Using feature/many-many-support as the base so that I could test with "full support" added. Plus there were to method name changes on that branch.

Manual test

I made a quick test that ran each build method 100 times, and then output the average time per iteration in milliseconds. I set this up on a recent project that has a reasonably small configuration layer (8 blocks, 2 pages, Site Config, etc).

I was surprised by how quick the build process was (on average) even when uncached, but it's worth mentioning that it will (of course) get worse with scale.

Result

Uncached - milliseconds per build iteration: 22.61
Cached - milliseconds per build iteration: 0.63

Test that I ran

public function testPerformance(): void
{
    // Instantiate Graph. This should trigger a build
    $graph = Graph::singleton();

    // Use ReflectionClass to access private methods
    $reflectionClass = new ReflectionClass(Graph::class);
    $build = $reflectionClass->getMethod('buildEdges');
    $build->setAccessible(true);
    $create = $reflectionClass->getMethod('buildGlobalCares');
    $create->setAccessible(true);

    // Start the test
    $start = microtime(true);

    $testCount = 100;

    // Run our builds
    for ($i = 0; $i < $testCount; $i += 1) {
        $build->invoke($graph);
        $create->invoke($graph);
    }

    $end = microtime(true);
    $milliseconds = ($end - $start) / $testCount * 1000;
    // Average time per build
    Debug::dump(round($milliseconds, 2));
}

Test coverage

• Test cache is available after build
• Test Graph properties are set correctly from cache on build
• Test Graph properties are set correctly from cache on secondary request

[adrhumphreys]

Left a comment regarding the terminology but it's not essential before merging this in

[adrhumphreys]

Can we change this to disallow list we should avoid using the terminology "blacklist"