[tracing] We run out of memory while taking a state snapshot in TF2

[jon-lunarg]

[plohrmann]

Things contributing to this problem:

1. TF2 hits memory limits on its own.
2. We're replaying the app and taking the snapshot in the same process space as vogleditor is running.

I think the best solution is to launch the replayer in a separate process and specify through cmd-line args at which API call the snapshot should be taken and where it should be saved (answer: in the current session folder). The Editor should then load the snapshot from disk.

This is in-line with a larger memory improvement that I've wanted to make - exactly such that most snapshots would be stored on disk and only the most recent 'x' snapshots would be kept in memory. Or maybe they are just loaded dynamically?

[plohrmann]

I just noticed that issue #113 is about launching the replayer in a separate process (there's no additional information in that issue though). I'm just indicating that they are related.