Shell integration: Explore how to handle sub-shells

[Tyriar]

I don't think we would be able to support sub-shells without a manual shell integration installation. Perhaps this should be an option for advanced users to be cwd/command aware even in sub-shells? There are other implications to this like knowing when the sub-shell starts and ends if we want to track the "tree". This would probably involve spelling out the shell ID (or SHLVL?) on every command.

[Tyriar]

We may get away with not bothering with a shell ID and instead of treating the buffer as a tree of commands, just compromise as a flat list where running a command to run a subshell is just treated as a single line. We definitely do need some way of differentiating machines though, other terminals use OSC 7 and include the hostname.

Some discussion in #145610 (comment)

[Tyriar]

The current thinking here is to document how to inject the script manually on the website to solve this more advanced case. The goal being users get a nice out of the box experience when shell integration is enabled and if users miss shell integration in sub-shells they can do the advanced setup which I don't think is possible without touching people's scripts which we'd rather not do.

[Tyriar]

Closing in favor of #151933 and microsoft/vscode-docs#5219 as manual install is the approach we'll recommend for sub-shell in the docs.