Move to ETS

[heywhy]

Overview

The changes contained in this PR allow the library to make use of ETS to store underlying indexed data. This improves performance and most importantly reduces the amount of data needed to be copied when the Elasticlunr.IndexManager.save/1 is called, as this is an obvious bottleneck.

Also, Task.async_stream/2 is now used when indexing documents because this allows us to analyze and process multiple documents concurrently.

TODO

• Decouple underlying index store to use ETS
• Adjust serialization/deserialization layer to work with these changes
• Run most indexing functions in an async stream
• Refactor the disk storage module
• GitHub Actions are all passing