Parse ConEmu OSC9

[mitchellh]

See: https://conemu.github.io/en/AnsiEscapeCodes.html#OSC_Operating_system_commands

Some other background: #3124, #3119

ConEmu OSC9 conflicts with iTerm notifications. We've supported iTerm notifications for a long time and I'd like to continue to do so. As a workaround, I think we should say that any ConEmu OSC9 commands can never be desktop notifications. This feels like a good tradeoff to me because why are you making those notifications anyways... and you can work around it by using OSC777.

We should expand our OSC parser to parse all ConEmu OS commands (OSC9).

As part of this, we should do better research on how ConEmu handles specific erroneous inputs because the table does not make it clear and there is no spec.

NOTE: This issue is only tracking parsing. We should not do anything with these commands. I'm not sure if we ever want to support some of these and for the others we should discuss in a GH discussion about what to do.

[Jan200101]

Here is how ConEmu implements the parsing https://github.com/Maximus5/ConEmu/blob/740b09c363cb16fbb730d72c53eaca1c530a016e/src/ConEmuCD/ConAnsiImpl.cpp#L2264
and here is how it handles the parsed data https://github.com/Maximus5/ConEmu/blob/740b09c363cb16fbb730d72c53eaca1c530a016e/src/ConEmu/RealConsole.cpp#L13656

notably it doesn't validate input beyond what it expects
so 9;4;3blablablabla isn't parsed further than 9;4;3

There is also the case that the "Running" state (9;4;1;pr) also has pr as optional component , despite what the documentation says, and defaults to 0

the Windows Terminal implementation is even more relaxed:
the OSC is segmented and parsed depending on the number of segments available

I think it might be reasonable to support a subset of what ConEmu does:

• be a lot more strict with the format, if anything that isn't expected is received fall back to desktop notification
• progress type can only be 1 char
• pr must be a number
• pr cannot be negative
• remove (0) doesn't support pr
• set (1) requires pr
• error (2) supports but doesn't require pr
• indeterminate (3) doesn't support pr
• pause (4) supports but doesn't require pr

this should follow the spirit of the documentation even if it doesn't follow the implementation of ConEmu or Windows Terminal

[mitchellh]

I should note that we don't need a single PR to do all of OSC9 at once to close this issue. In fact, I'd prefer that not the case. I'd rather see small scoped PRs so do less, add tests, make it clean. ❤

And thanks @Jan200101 this is great information!