q_overlnd_x and q_overlnd_y were not being initialized if surface_predictor == True but are used in the surface predictor
calculation even if they are not output. This was causing PF to throw an error for this case.

---------

Co-authored-by: Steven Smith <smith84@llnl.gov>

### Summary

This PR standardizes CMake variable naming and updates GPU
documentation. It addresses a breaking configuration change where
`RMM_ROOT`, originally defined in uppercase, was later changed to
mixed-case (`rmm_ROOT`), which silently broke existing user build
scripts when upgrading ParFlow. This PR restores the original uppercase
convention and enforces consistent naming across all third-party
dependencies to prevent similar regressions.

### Changes

- **CMake variable consistency:**
  - All third-party CMake variables are now uppercase, including:
    - `RMM_ROOT` (restored from mixed-case `rmm_ROOT`)
    - `UMPIRE_ROOT` (updated from `umpire_ROOT`)
- This aligns with existing variables such as `KOKKOS_ROOT`,
`SILO_ROOT`, `HYPER_ROOT`, and `SUNDIALS_ROOT`.
- **Workflow files** updated to reflect the uppercase RMM_ROOT and
UMPIRE_ROOT variables.
- **Backward compatibility fix:**
- `RMM_ROOT` was originally uppercase; changing it to lowercase silently
broke existing user scripts when upgrading to newer ParFlow versions.
This PR restores the original convention to avoid unexpected build
failures.
- **Documentation updates (README-GPU.md):**
- Removed obsolete references to RMM v0.10 and the previous upgrade
work-in-progress note
  - Ensured consistent use of `RMM_ROOT`
  - Updated the RMM link to the official repository

## Summary
Add configurable snow parameterization options for CLM.
## New Features
- **Rain-snow partitioning:** CLM, WetbulbThreshold, WetbulbLinear
methods
- **Thin snow damping:** Prevents spurious early-season melt from warm
ground
- **Albedo schemes:** CLM, VIC, Tarboton temperature-dependent aging
## New Keys (12 total)
- `Solver.CLM.SnowPartition`, `WetbulbThreshold`
- `Solver.CLM.ThinSnowDamping`, `ThinSnowThreshold`
- `Solver.CLM.AlbedoScheme`, `AlbedoVisNew`, `AlbedoNirNew`, `AlbedoMin`
- `Solver.CLM.AlbedoDecayVis`, `AlbedoDecayNir`, `AlbedoAccumA`,
`AlbedoThawA`
## Changes
- YAML schema with types, defaults, and bounds
- RST documentation with physics background
- 3 bibliography references
- 3 Python tests with reference outputs
## References
- Wang et al. (2019) GRL, doi:10.1029/2019GL085722
- Tarboton & Luce (1996) WRR, doi:10.1029/96WR01049
- Andreadis et al. (2009) JHM
## Test Plan
- [x] `clm_snow_defaults` - backward compatibility
- [x] `clm_snow_partition` - WetbulbThreshold method
- [x] `clm_snow_albedo` - Tarboton scheme
  All pass via `ctest -R clm_snow`

---------

Co-authored-by: Reed <rmaxwell@mines.edu>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Steven Smith <smith84@llnl.gov>

## Summary
Extends CLM snow parameterizations with new rain-snow partitioning
methods, SZA-based melt damping, and configurable fractional snow cover.
**New Features:**
- **Dai (2008) rain-snow partitioning**: Hyperbolic tangent function
fitted to 30 years of global observations
- **Jennings (2018) rain-snow partitioning**: Bivariate logistic
regression using temperature and relative humidity
- **SZA-based snow melt damping**: Reduces melt energy at high solar
zenith angles where CLM underestimates albedo
- **Configurable fractional snow cover**: Adjustable roughness parameter
for frac_sno calculation
- **Configurable temperature thresholds**: SnowTCrit, SnowTLow,
SnowTHigh, SnowTransitionWidth for existing methods
**New Keys (16 total):**
- `Solver.CLM.SnowTCrit`, `SnowTLow`, `SnowTHigh`, `SnowTransitionWidth`
- `Solver.CLM.DaiCoeffA`, `DaiCoeffB`, `DaiCoeffC`, `DaiCoeffD`
- `Solver.CLM.JenningsCoeffA`, `JenningsCoeffB`, `JenningsCoeffG`
- `Solver.CLM.SZASnowDamping`, `SZADampingCoszenRef`,
`SZADampingCoszenMin`
- `Solver.CLM.FracSnoScheme`, `FracSnoRoughness`
**References:**
- Dai (2008) GRL doi:10.1029/2008GL033295
- Jennings et al. (2018) Nature Communications
doi:10.1038/s41467-018-03629-7
- Dang et al. (2019) The Cryosphere doi:10.5194/tc-13-2325-2019
## Test plan
- [x] `clm_snow_dai` - Dai sigmoidal partitioning method
- [x] `clm_snow_jennings` - Jennings bivariate logistic method
- [x] `clm_snow_sza_damping` - SZA-based melt reduction
- [x] All existing CLM snow tests pass
Run tests with: `ctest -R clm_snow`
  🤖 Generated with [Claude Code](https://claude.ai/code)

---------

Co-authored-by: Reed <rmaxwell@mines.edu>
Co-authored-by: Claude <noreply@anthropic.com>