Simple changes to handle missing values in BayesianModel predict methods

[robbymeals]

There may be a better/faster way to do this, but supporting missing variables easily in these predict methods will definitely make my own life easier.

Fixes #1118 . (if there is no issue for this, please create one)

Changes

• Checks if input state name is either None or nan before converting to state index, returns original value if it is instead of failing from KeyError
• Drops NA from series in map_query calls over iterrows

Thank you!

[codecov]

Codecov Report

Merging #1119 into dev will decrease coverage by 0.00%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##              dev    #1119      +/-   ##
==========================================
- Coverage   92.93%   92.93%   -0.01%     
==========================================
  Files         130      130              
  Lines       12632    12634       +2     
==========================================
+ Hits        11740    11741       +1     
- Misses        892      893       +1     

Impacted Files	Coverage Δ
pgmpy/models/BayesianModel.py	88.27% <ø> (ø)
pgmpy/utils/state_name.py	77.14% <66.66%> (-1.65%)	⬇️

---

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 b2ea305...f48c543. Read the comment docs.

[ankurankan]

@robbymeals Thanks for the PR. I don't know much about different ways of dealing with missing values in BN and will have to do some literature reading for that.

For your changes could you add a warning message before dropping any rows in the data so that the user is aware of it? Also, could you please add some tests for the changes?