perf: improve creating tfIdf with cache

[vivekjoshi556]

Description

Improve #353

While building indexes there is a call to creating TfIdf.
I noticed that the the getIdf() method only depends on the word and doesn't the other values remained constant, so I cached the result of getIdf() for different words and before calling it again I check if I already had the value.

This improved the performance significantly.

getTfIdf() performance:
Before: 1:52.468s. (maybe my sys is slow)
After: 9.021s

Checklist

Please review this checklist before submitting a pull request.

• [x ] Code compiles correctly
• Created tests, if possible
• All tests passing (npm run test:all)
• Extended the README / documentation, if necessary

[vivekjoshi556]

As mentioned in #350 . @agnivade you mentioned:

I think it's rare to see this in a practical scenario. You want to lookup a command, you see that it doesn't exist. Would you want to retry that command again in 2 minutes? That seems unlikely. But I agree that there's definitely things to improve on. Let's track those here: #353

I agree that same command might not be looked up immediately but what if user looks for some other command that also doesn't exist.
If you agree I would like to work in this idea as well and add this suggestion in this commit as well.

[kbdharun]

LGTM, Thanks for this enhancement (and sorry for the late review). Tested it on Windows and Linux the indexing in the update process is much faster now.