Major update for modern neuralhydrology configuration and ensemble predictions

[jmframe]

This pull request fixes quite a few problems. Sorry for the major PR, but I'm not always sure how to do these things. The two big reasons for this PR is 1) including trained LSTM weights with AORC forcings that were trained with the new scalar configuration from neuralhydrology, and 2) adding the ability to make ensemble mean predictions from the LSTM BMI.

Additions

• capability to make ensemble mean predictions from multiple trained LSTM weights. Described in this issue: LSTM predictions from ensembles of training initializations #53
• BMI configuration files for over a dozen examples that can run with the demonstration notebooks.

Removals

• Some outdated configuration files
• outdated trained LSTM models addressing issue remove out of date example model weights and run notebooks #52
• outdated notebooks that aren't compatible with new weights and BMI.

Changes

• scalar values for processing model input and output data now come from a yaml file, described in this issue: scalar values from trained neuralhydrology model can be updated to avoid pickle. #51.
• the sample model weights in the trained_neuralhydrology_models directory now come from the updated neuralhydrology repository, and now are trained on AORC forcing data hosted publicly here: https://doi.org/10.4211/hs.c738c05278a34bc9848dd14d61cffab9
• the file to generate sample configuration files has been updated to include multiple trained LSTM weights.

Testing

1. Tests were done to ensure the BMI successfully runs and the LSTM produces good predictions with ngen compiled on Pantarhei
2. Tests were done to ensure the BMI successfully runs and the LSTM produces good predictions with NGIAB
3. Two demonstration notebooks show the successful module runs with sample data in Jupyter Notebooks: https://github.com/jmframe/lstm/tree/readme_update/notebooks

[JoshCu]

Tested using nextgen in a box with both a single model and ensemble for gage-14216500 --start 1998-11-01 --end 1999-01-01 Inputs generated using this tool / branch
Single model

train_cfg_file:
 - /ngen/ngen/extern/lstm/trained_neuralhydrology_models/nh_AORC_hourly_25yr_1210_112435_7/config.yml

Ensemble

train_cfg_file:
  - /ngen/ngen/extern/lstm/trained_neuralhydrology_models/nh_AORC_hourly_25yr_1210_112435_7/config.yml
  - /ngen/ngen/extern/lstm/trained_neuralhydrology_models/nh_AORC_hourly_25yr_1210_112435_8/config.yml
  - /ngen/ngen/extern/lstm/trained_neuralhydrology_models/nh_AORC_hourly_25yr_1210_112435_9/config.yml
  - /ngen/ngen/extern/lstm/trained_neuralhydrology_models/nh_AORC_hourly_25yr_seq999_seed101_0701_143442/config.yml
  - /ngen/ngen/extern/lstm/trained_neuralhydrology_models/nh_AORC_hourly_25yr_seq999_seed103_2701_171540/config.yml
  - /ngen/ngen/extern/lstm/trained_neuralhydrology_models/nh_AORC_hourly_slope_elev_precip_temp_seq999_seed101_2801_191806/config.yml