L2metrics

- Implemented variant-aware and variant-agnostic metric calculations
- Changed default STE averaging method to metrics
- Changed order of pre-processing to smooth then clamp then normalize
- Included block subtype in regime metrics dataframe
- Handled exception occurring with tasks in STE data without normalization ranges
- Added dashed line on STE plots where performance curves are stitched together
- Added feature to plot all STE runs when averaging with metrics method
- Added feature to calculate runtime from log timestamps
- Updated README and example output files

- Added option to evaluate a single agent configuration directory
- Added more data validation for normalization and loading STE data
- Added support for M15 evaluation directory validation
- Implemented flag for enabling/disabling evaluation block data smoothing
- Redirected console output to tqdm write when showing progress bars

- Consolidated settings output into one JSON file
- Added normalization range in metrics JSON for each run
- Disabled smoothing for evaluation block data
- Fixed bug with regime and block number mismatch by filtering NaNs before filling regime number
- Implemented using input data range for clamping outliers
- Updated normalization range to 1-101
- Implemented new block plotting visualization
- Created utility function for initializing l2metrics parser
- Fixed bug with numpy array type when aggregating by median
- Removed deprecated command-line argument: --show-rate-data

- Updated evaluation package for M12 protocol
- Replaced lifetime aggregation method calls with numpy version to ignore NaNs
- Fixed bug with time-series averaging for multiple STE runs
- Added checks for proper settings before creating output directories
- Updated evaluation block plotting method to show pre-processed data instead of just mean line

- Applied recursive option for storing STE data
- Aggregated log data into single feather file for recursive flag
- Output plots for recursive flag
- Add output directory flag to l2metrics
- Updated documentation and parameter descriptions
- Updated metrics calculations for scenarios with wake/sleep blocks
- Fix bug with learning block length of 1 when computing recovery time

- Updated reference to metrics specification document 0.68
- Updated default settings for aggregation (mean) and transfer (ratio)
- Fixed bug with setting normalization method to none
- Clamp outliers in STE data as well as LL data