Most gapit verbs can now take a capture ID instead of a gfxtrace file.

[paulthomson]

I would like to be able to pass a capture ID and a gapis-port to gapit screenshot so that it will take a screenshot of some capture that is already loaded. This avoids me having to wait 5 mins per screenshot to load the capture; instead, I can load it once from disk and then refer to it later via its capture ID. I would also like to modify (i.e. create mutated versions of) captures in gapis without writing them to disk, so this PR will be useful in this scenario.

Slightly weird: since the capture file is a positional/non-flag argument, I am attempting to parse it as a hex string (capture ID) and, if that fails, treating it as a file as before. If you have a filename that happens to be a hex string, you could use e.g. "./0000FFFFF". Could alternatively use a flag and then expect no positional args.

[AWoloszyn]

You could add a boolean flag to gapit, (--capture-id) that just tells gapit that the given string is a capture-id instead of a file.

We have several boolean flags already, and this would make it more unambiguous. (-noopt for example).

[paulthomson]

Talked with Ben. Made most gapit verbs be able to take a capture ID. If you add -captureid, the capture arg will be interpreted as a capture ID instead of a capture file path. Most verbs get gapis and load a trace, so created a getGapisAndLoadCapture function.

[paulthomson]

I also added another change: gapit replace_resource can optionally not write out a capture file by passing a flag: -skipoutput. This allows me to mutate traces in gapis and then get screenshots from them, without having to write them to disk.