Adding Utility to Detokenize as list of Strings to Tokenizer Base Class

[aflah02]

Fixes #113
Opening this PR so that we can discuss the name of the utility
I'll write tests as discussed in the issue and fix some problems currently in the meanwhile

[aflah02]

@mattdangerw Since I also need to test the Unicode tokenizer all it's commits will reflect here too. Should I close the other PR then?

[aflah02]

@mattdangerw The PR is ready for review.
There is an issue with the tests with ragged values but when I run it simply on colab nothing happens and it works as intended as can be seen here however the test fails and I suspect it is due to the warning which is raised due to some other components which I haven't tinkered with :

[mattdangerw]

@aflah02 sorry missed this comment. We should land this as a follow up to the other PR. Can you rebase this so we can review just this change?

Also I would add your tests to the base tokenizer as a that's where the functionality is. You may need to make a new simple subclass of the Tokenizer base class in the unit test.

Re that error, I am not sure what is going on there, it may be a incompatibility between tensorflow and numpy when you call ragged.to_list(). If so we probably don't need to fix that here.

[aflah02]

@mattdangerw
No Worries!
Sure I'll do that, also I might refork and make the changes there as I messed up my branch locally and need to fix that as well. I'll open a new PR then and reference this