CEED_STRIDES_BACKEND related updates

[jeremylt]

I have two four changes here

1. Expanded note in the documentation about CEED_STRIDES_BACKEND

2. Set CEED_STRIDES_BACKEND to be different for memcheck than the default layouts for both CPU and GPU

3. Add CeedElemRestrictionGetLLayout for strided restrictions

4. Fix CeedElemRestrictionGet[L,E]Layout signature, previously there ways an unneeded pointer

This won't provide a clear warning message, but it will at least demonstrate bugs if a user mixes and matches memcheck CEED_STRIDES_BACKEND with other backends.

I don't think we can be really any clearer with the name that CEED_STRIDES_BACKEND is specific to the backend?

Offering CeedElemRestrictionGetLLayout probably fixes the root cause of why someone would see this problem though.

See #1460

Note: if we like this, a666b53, 0e6bd4c, and 32058e1 should be squashed together

Also note: if we like CeedElemRestrictionGetLLayout, I'll go add some quick unit test

[jedbrown]

LGTM at a high level. Thanks for solving this problem. I take it we aren't enforcing anything new outside the memcheck backend?

[jeremylt]

I take it we aren't enforcing anything new outside the memcheck backend?

The only thing that is disruptive is f447f4b - really this function signature should have never had a pointer to an array. Its not a big change for users and I think we should fix.

We aren't forcing any changes in how the current API behaves past that.