Add Cmd::before_spawn to allow modifying the resulting std::process::Command before spawning it

[swlynch99]

Hey! I've been using this crate a bunch and it's been great!

For one of my use cases I need to run a process under a different uid. If I want to do this I can use xshell to build the command then I need to eject out and convert it directly to std::process::Command to set the uid. If I do that, though, I can no longer use read and I end up needing to write a bunch of code to replicate it.

Adding a uid method felt too platform-specific so this PR instead adds a before_spawn method that registers a function that gets called with the std::process::Command just before it is spawned. It is based on the before_spawn method in duct which does the exact same thing.

Notes

• CmdData needs to be Clone + Send + Sync and Box<dyn Fn(&mut Command)> is none of those so I ended having to place the list of methods directly in Cmd.
• I started off trying to bound the functions by 'a but that broke one of the tests so I've left the hook as Box<dyn Fn(&mut Command) + 'static> instead.
• Hook functions don't get run when converting Cmd to std::process::Command since they are fallible. I have documented this in the method documentation.

[matklad]

Hm, this is definitely a valid use-case, but I think I prefer this to be out of scope for xshell. With 0.3.0, the recommended way is indeed to go via conversation to Command path. While more code to write, it keeps the upstream API simple, and I value that somewhat more!

Thanks for the PR anyway, and sorry for ridiculously long response time!