Calibrate global bucket LHF and SHF

[AlexisRenchon]

closes #871

This PR goal is to build a global calibration pipeline (minimal working example - MWE) to calibrate a land model globally.
In this commit, we calibrate a bucket model (the framework for full land model would be the same).
Two different approaches are added to this commit: the same calibration is done with just EnsembleKalmanProcesses.jl - EKP, or with ClimaCalibrate.jl (ClimaCalibrate.jl uses EKP, but has different backend, optimised for different HPCs).
In our MWE, we use ERA5 latent heat - LH and sensible heat - SH flux as target observations for the calibration. We stack monthly observation at n (e.g., 50) locations on land.
We calibrate the following parameters with following priors:

prior_κ_soil = EKP.constrained_gaussian("κ_soil", 2, 1, 0, Inf);
prior_ρc_soil = EKP.constrained_gaussian("ρc_soil", 4e6, 2e6, 0, Inf);
prior_f_bucket = EKP.constrained_gaussian("f_bucket", 0.5, 0.3, 0, 1);
prior_W_f = EKP.constrained_gaussian("W_f", 0.4, 0.4, 0, Inf);
prior_p = EKP.constrained_gaussian("p", 2, 1, 1, Inf);
prior_z_0m = EKP.constrained_gaussian("z_0m", 0.01, 0.1, 0, Inf);

and we use:

ensemble_size = 10
n_iterations = 5

Future work will use this pipeline to calibrate the land model globally.

TEMPORARY NOTE - delete before merge:

• EnsembleKalmanProcesses.jl - done!
• ClimaCalibrate.jl Script testing WIP... error: NetCDF error: HDF error. From Gabriele: "I see this error when NC files are already open somewhere, for example, if it is still being written. Make sure that writer are closed when the workers are done"