Bring in FTorch

[samsrabin]

Tried to do this in ctsm5.3.051, but it didn't include the right CMEPS version; see #3214.

@adrifoster reports a lot of additional work is required to actually use FTorch with CTSM.

Definition of done:

• FTorch added as a submodule
• Any scripts/etc. needed for an FTorch workflow are added to CTSM
• Any required Python packages are added, if needed, to ctsm_pylib
• Workflow documentation added to User's Guide
• At least one FTorch-using test is added to aux_clm

[wwieder]

Unless this is needed sooner, I suggest we give this a CESM3.1 milestone and will mark as low priority (for now).

[dlawrenncar]

For this, I think the important thing is that the @linniahawkins and LEAP / M2LINES groups have something to work with. Based on my conversations with Linnia, it seems like they do, but that should be confirmed

[linniahawkins]

Adrianna got us set up so this is low priority for now, Thank you.

[samsrabin]

@adrifoster and @linniahawkins: In the spirit of this article (document and share as much as possible), would it be possible to get that workflow posted somewhere? Maybe on the GitHub wiki?

[jedwards4b]

I just ran across this when looking for another issue - this should be straight forward to do if you start from the work I have done for cam in:
https://github.com/jedwards4b/CAM/tree/cam6_4_093_ftorch. please don't hesitate to ask if you have any questions or concerns - I am really trying to coordinate so that everyone can use the same interface to ftorch.

[samsrabin]

Pinging this; @linniahawkins only ran into #3552 because we don't have FTorch in CTSM (thus she did a full CESM checkout). I think we should try to bring this in by the release, since FTorch is actively being used already, and Jim suggests it shouldn't be too much work.