Broad XX enhancement and performance improvements

[mmaiers-nmdp]

• Closes DRB1*02:XX #31 creates mappings for broad XX allele codes like DRB1*02:XX based on static file
• pyard/dna_relshp.csv added to repo. (extracted from STAR mdp_cmn_prd..t_dna_relshp) no changes in 20 years)
• Modified logging to not reconfigure format.
• added documentation
• Several performance enhancements

1. refactored isvalid() function to use a dictionary instead of testing whether allele is in a list
2. refactored redux() function to use a dictionary instead of testing whether allele is in a list
3. precompiled regex instead of compiling it during each call to redux() for performance
4. used memoization to store results of previous calls to redux()

profile of convert_pull.py before performance enhancements

   744150  254.186    0.000  255.649    0.000 pyard.py:457(isvalid)
   371361  107.522    0.000  108.164    0.000 pyard.py:366(redux)
  5227562    9.539    0.000    9.979    0.000 pyard.py:51(getvalue)
        1    6.400    6.400    6.400    6.400 {built-in method _pickle.load}
  2613781    4.156    0.000   16.305    0.000 pyard.py:55(loci_sort)
  7501937    3.821    0.000    5.614    0.000 re.py:271(_compile)
  7501306    3.100    0.000   11.050    0.000 re.py:180(search)
 10968273    2.467    0.000    2.467    0.000 {method 'split' of 'str' objects}
  7501361    2.360    0.000    2.360    0.000 {method 'search' of 're.Pattern' objects}
     6020    2.207    0.000    2.207    0.000 pyard.py:450(<lambda>)

profile of convert_pull.py after performance enhancements

   281302   89.124    0.000   89.349    0.000 pyard.py:384(redux)
  5237286    9.816    0.000   10.269    0.000 pyard.py:52(getvalue)
        1    5.149    5.149    5.149    5.149 {built-in method _pickle.load}
  2618643    4.335    0.000   16.832    0.000 pyard.py:56(loci_sort)
     1439    3.616    0.003    3.616    0.003 {method 'flush' of '_io.TextIOWrapper' objects}
  7130577    3.144    0.000    4.640    0.000 re.py:271(_compile)
  7129945    2.758    0.000    9.248    0.000 re.py:180(search)
 10987721    2.468    0.000    2.468    0.000 {method 'split' of 'str' objects}
     6020    2.350    0.000    2.350    0.000 pyard.py:473(<lambda>)
  7411302    1.943    0.000    1.943    0.000 {method 'search' of 're.Pattern' objects}

the improvement in isvald() has been dramatic.
the improvement in redux() has been less substantial.

further performance enhancement suggestions

• increase memoization cache size for redux(). Currently using @functools.lru_cache(maxsize=1000)
• add memoization to getvalue()
• have a code clinic to discuss other ideas (parallelize? more caching/memoizing?)

[cicd-nmdp]

🥇 LGTM

[pbashyal-nmdp]

LGTM 👍