LibWeb: Implement navigator.deviceMemory

[gmta]

Fixes at least 2 subtests in wpt/device-memory.

[awesomekling]

Mild fingerprinting concern, but it's not like we have a structured way of dealing with that at this point in time.

[gmta]

Mild fingerprinting concern, but it's not like we have a structured way of dealing with that at this point in time.

The spec already tries to prevent fingerprinting somewhat by rounding to a power of 2 (which, of course, does not help a lot) and clamping the resulting value. We could tighten up the lower and upper bounds of the values we return; that's allowed by the spec.

[gmta]

I went ahead and tightened the clamps on the returned values. It now only returns 1, 2 or 4 GiB of physical memory. I expect this to be enough for code to determine whether they're dealing with a very low memory device or one with "enough" memory, while improving fingerprinting prevention a bit.