Ensure search results include all parts-of-speech

[rektdeckard]

This patch fixes a bug in which loading a lemma from the database would clobber any existing entries to the database from a different part of speech but with same lemma. For example, the adjective sense of the word "smart" ("showing mental alterness...") was overwritten by the noun sense ("a kind of pain such as that caused by a wound...") when the data.noun file was loaded on init().

The Database now stores lemmas as a Map<string, Map<string, Index>>, where the outer key is the lemma itself, pointing to a map of POS -> Index. In this way, you can access the requisite POS by calling indexLemmaIndex.get("smart")?.get("adjective");.

This affects the return signatures of searchFor() and searchSimpleFor(), and should be considered a major semver update.

[rektdeckard]

Apologies for all the formatting changes! I can resubmit without them if it's a problem.

[sudhanshuraheja]

Thanks @rektdeckard. Let me go through this now.

[rektdeckard]

0c7758c also should solve #2, so you only need to:

const Dictionary = require('en-dictionary');

When making use of the library.