Fix CMake Variable Naming

[wavefunction91]

Previously, the GAUXC_ENABLE_XYZ variables did all the heavy lifting for feature detection and control. Per @loriab's comments and subsequent issue (#102),these variables sound suspiciously editable, so probably best to not rely on that naming scheme.

This PR separates out GAUXC_ENABLE_XYZ to trigger detection of feature XYZ and sets GAUXC_HAS_XYZ if the feature has been detected and enabled throughout the stack.

Major Change: Do not use GAUXC_ENABLE_XYZ downstream to check feature behaviour, only use these variables to enable features via FetchContent and use GAUXC_HAS_XYZ to check if the feature enable happened. The latter should also be used for source-level feature detection (e.g. GAUXC_HAS_DEVICE can be used to check if GauXC has device bindings)

Closes #102

TODO: Check downstream integrations:

• NWChemEx
• MPQC
• ExaChem
• ChronusQ
• Psi4

[wavefunction91]

Ping @evaleev @dmejiar @ajaypanyala @elambros @loriab @davpoolechem

This will undoubtedly break some things in some build systems. Please check that you can make this work in your respective code bases. I won't merge until we can get consensus that we're not breaking things downstream.

[ajaypanyala]

ExaChem is good to go!

[wavefunction91]

MPQC is fixed as of https://github.com/ValeevGroup/mpqc4/pull/464 @evaleev.

[davpoolechem]

Can confirm that this works for Psi4!

[wavefunction91]

NWChemEx integration is fixed

[wavefunction91]

@elambros can you confirm that this PR works in CQ? It will require changing all instances of GAUXC_ENABLE_DEVICE to GAUXC_HAS_DEVICE for the device feature check in the code (fetch/configure should stay the same)

[elambros]

Yep works on the CQ end too!