Enable consumption of partially written audio mid-stream

[jstangroome]

While one workflow continues to write an ongoing stream of frames to the encoder, enable the file and chunk size headers to be updated so that a valid snapshot can be taken and then continue to write more frames to the encoder.

This is a backward compatible change that should have no impact on use cases that do not need this capability. Concurrency concerns are left to the caller, along with efficient methods to snapshot the data before additional writes continue.

My use case is to repeatedly pass snapshots of the audio-so-far to a third-party API for analysis while new audio continues to be written.

This would likely also enable #36 to be solved by the caller.

Importantly any custom metadata is still only written upon Close().

While this is mostly a refactor of existing code within Close to be callable independently of Close I did also replace some magic numbers with more verbose definitions.