Resolve PCSS issues, remove "experimental" flag, and "release" the feature

[JMS55]

Bevy version

0.15

Info

• Calculate the search radius based on the light radius in world units and the distance to the light in world units (for directional lights this should be configurable and perhaps default to the distance to the sun)
• Average depth in world units when finding the blocker
• Use the correct calculation for the filter radius that leverages the penumbra size calculation
• Use blue noise random rotations of the MSAA sampling pattern

From this conversation on discord https://discord.com/channels/691052431525675048/743663924229963868/1300819753480228895

[cart]

Given the issues with the implementation, I think we should make PCSS experimental for 0.15 and hold off on announcing it until we resolve these problems.

@pcwalton does that sound reasonable?