Add support for "float32-filterable" feature.

[HymnsForDisco]

Is your feature request related to a problem? Please describe.
Without this feature, float32 type textures don't support sampling. Practically, this means you must use FilterMode::Nearest for samplers of such textures, which is rather limiting.

Describe the solution you'd like
I'd like to be able to enable the "float32-filterable" feature in the device descriptor

Makes textures with formats r32float, rg32float, and rgba32float filterable.

Presumably, this would join the other features enumerated in https://docs.rs/wgpu/latest/wgpu/struct.Features.html

Describe alternatives you've considered
The alternatives AFAIK are:

1. Don't use float32 textures when you want sampling
2. If you need to sample from a float32 texture, first copy/convert it into a sample-able format. (This may be workable in some cases, but bad for frequently updating textures, and doesn't support the full range of values that float32 textures do).

Additional context
I originally asked about this feature here, before deciding a feature request was appropriate.
#4846

[teoxoy]

This landed a few days ago in #4759.