Cleanup filtering code and behavior in preprocess

[Lioscro]

I noticed there were a few inconsistencies between the variable name & docstring vs the actual filtering behavior in pp.utilities.filter_cells and pp.utilities.filter_umis. This PR addresses these inconsistencies.

• Renamed the umi_read_thresh argument to pp.filter_molecule_table to min_reads_per_umi to match the naming of the other variables. The corresponding argument to pp.utilities.filter_umis has also been changed.
• All filter thresholds (min_umi_per_cell, min_avg_reads_per_umi, min_reads_per_umi) specifies the minimum number required for a cell/UMI to pass filtering. Previously, when a condition equaled any one of these values, the cell/UMI would be filtered out (due to the use of a <= operator.
• Removed the second (redundant) filtering step that was not actually performing any additional filtering in pp.filter_molecule_table.
• Updated filtering code in pp.utilities.filter_cells to perform pandas operations more efficiently by using groupby operations.

Overall, these changes result in around 10x speedup of the pp.filter_molecule_table function.

[codecov[bot]]

Codecov Report

Merging #137 (af4997e) into master (53e705b) will decrease coverage by 0.04%. The diff coverage is 82.85%.

@@            Coverage Diff             @@
##           master     #137      +/-   ##
==========================================
- Coverage   86.28%   86.24%   -0.05%     
==========================================
  Files          65       65              
  Lines        4536     4530       -6     
==========================================
- Hits         3914     3907       -7     
- Misses        622      623       +1     

Impacted Files	Coverage Δ
cassiopeia/preprocess/cassiopeia_preprocess.py	23.07% <0.00%> (ø)
cassiopeia/preprocess/constants.py	100.00% <ø> (ø)
cassiopeia/preprocess/UMI_utils.py	95.25% <50.00%> (-0.42%)	:arrow_down:
cassiopeia/preprocess/pipeline.py	71.01% <77.77%> (-0.16%)	:arrow_down:
cassiopeia/preprocess/utilities.py	96.56% <100.00%> (-0.07%)	:arrow_down:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 53e705b...af4997e. Read the comment docs.