Read file as Buffer to mitigate V8's 256MB max string size

[JamesMGreene]

Read file as Buffer, read lines as strings.

Implementation concept can be taken from existing NeDB PR: https://github.com/louischatriot/nedb/pull/493

However, it's not a terribly performant implementation since it is doing things like concatenating an array with itself rather than just pushing new items on to it, so a rewrite is in order.

This would mitigate frustrating issues like the following:

• https://github.com/louischatriot/nedb/issues/363
• https://github.com/louischatriot/nedb/issues/389

[JamesMGreene]

This looks like a better implementation to consider: https://github.com/louischatriot/nedb/pull/463

[JamesMGreene]

Seems like a symmetrical method for writing as a stream would also be necessary for the same reason.

[JamesMGreene]

The default writeAsStream implementation should be very mindful of honoring backpressure cues from Node like in the examples from the documentation and other sources, e.g. this [imperfect] StackOverflow thread.

[JamesMGreene]

IMPORTANT NOTE FOR BROWSER COMPATIBILITY

All Stream-specific module usage should be limited to "storage.js" (which is subbed out for the browser version, for example) and NOT put into "persistence.js" (which is NOT subbed out).

It is acceptable to have stream-related work in "persistence.js", e.g. listening for events, so long as it

• is not required (i.e. still support the non-streaming version)
• does not require additional module dependencies specific to streams

[JamesMGreene]

For reference: https://github.com/nodejs/node/issues/3175