Fix input sampling validation and SIPNET IC handling

[blesson-07]

Issue Title
Fix Input Sampling Validation and Deterministic IC Handling

Problem
PEcAn improperly passed multiple input paths (e.g., initial conditions) to model configuration functions without upstream sampling. Models like SIPNET handled this by randomly selecting a path, violating PEcAn’s design where sampling should occur at the ensemble level. This led to:

Non-reproducible workflows (random IC selection).

Lack of validation for unsampled multi-path inputs.

Solution
Upstream Validation:
write.ensemble.configs now errors if an input has multiple paths but no sampling method ( in XML).
Ensures only one path per input is passed to model configs.

[blesson-07]

@mdietze @infotroph
Could you review the input validation changes?

[infotroph]

Thank you! Overall this is looking very good. I'll leave some notes inline, but the main change I see needed is to put the test file in modules/uncertainty/tests/testthat rather than base/all/[...]. I also see what looks to me like the same check in two places -- is that a true duplicate or am I not seeing the reason both are needed?

[infotroph]

@blesson07asd Will you be able to return to this soon? Seems like it's in pretty good shape overall.

The biggest issue I see remaining is that ensemble-test.R is now removed entirely -- thanks for taking it out of base/all as requested, but can you please add it back in modules/uncertainty?

[blesson-07]

@infotroph Apologies for the delay, I was in the middle of my semester exams, which are now over. I’ll resume work on this and get the remaining changes pushed ASAP. Thanks for your patience and support!

[mdietze]

Not sure why you added rhdf5 to the libraries in data.remote as you don't seem to use it anywhere in your code