EAMxx: refactor vertical remapper to handle different top/bot extrapolations

[bartgol]

In view of unifying the "dataset interpolation" implementations across eamxx, this PR makes the VerticalRemapper a bit more general. Namely:

• can be constructed from two existing grids (as opposed as assuming tgt grid is from map file);
• can handle "P0" and "Masked" extrapolation (the former does constant extrapolation, the latter sets extrapolated value to a constant mask value)
• can do different extrapolations at top and bottom of the column (but same extrapolation type across all columns)
• can handle 1d as well as 2d pressure profile, for both source and target pressure

Depends-On: #6761

[github-actions]

PR Preview Action v1.4.8
🚀 Deployed preview to https://E3SM-Project.github.io/E3SM/pr-preview/pr-6770/
on branch gh-pages at 2024-11-26 22:39 UTC

[rljacob]

@AaronDonahue or @tcclevenger please review.

[bartgol]

@tcclevenger is on the road, moving across country.

@AaronDonahue, the CUDA fails are the same as in master. The gh/ci fails are mysterious, but maybe @mahf708 knows what the deal is. That said, it's likely orthogonal to this PR.

[bartgol]

(a)ccidentally deleted the branch)

[bartgol]

The last round of testing showed all passing and the known master diff on CUDA. The ghci-snl runners are having network issues right now, so they are stuck, but nothing changed since the last round of tests, so I'm merging.