Drop unused serialised field

[rolftimmermans]

This drops df from the indexed data and stores only the frequencies per document ID. Serialisation version is bumped to 2. I have not added any code to accept version 1 dumps too, but it should be fairly easy to add if this is desired.

This PR is based off of #135 for now.

Before

Index size: 13497 terms, 14097 documents, ~15.49MB in memory, 2.49MB serialized.

Load index:
===========
  * MiniSearch.loadJSON(json, options) x 21.84 ops/sec ±3.51% (41 runs sampled)

After

Index size: 13497 terms, 14097 documents, ~15.72MB in memory, 2.31MB serialized.

Load index:
===========
  * MiniSearch.loadJSON(json, options) x 23.81 ops/sec ±3.35% (44 runs sampled)

[lucaong]

I merged #135 , squashing commits, could you rebase this one when you have time?

[rolftimmermans]

Done!