write peerconnections to different files

[fippo]

currently, all data associated with a session is written to a single file. Splitting this up into one file for session metadata and another file for each peerconnection would allow processing a peerconnection once its done and not when the client closes the page.
Increases realtimeness and hopefully also reduces disk usage.

[fippo]

also this would allow processing a connection once its done without waiting for the websocket to be closed. This will however change how you download and inspect dumps -- the current model of one-big-file is pretty nice but that could be retained with the right tooling.

[dagingaa]

This would be cool for our sake, as long as it also triggers feature extraction for that peer connection. But I respect that this might be worse in a multiparty-setting

[johnryan]

@fippo I took a stab at this in #275, let me know if you have any feedback!

[johnryan]

@fippo I came back to check on this (time flies 😬) and see if there was any interest in this feature anymore - looks like my PR is out of date. I'd be happy to update it if it will get merged but otherwise happy to close the PR.