Avoid leaking inner error types

[jessebraham]

See the "Common mistake: inner error type leakage" section of this blog post for more information:
https://d34dl0ck.me/rust-bites-designing-error-types-in-rust-libraries/index.html

Given the relatively small amount of effort required, I think it's worth doing this. We can create a type alias for the Box< .. > mess and just use that everywhere that we are currently leaking error types.

This is only applicable to library errors, but may be worth just doing everywhere. Less concerned about the CLI errors.

[jessebraham]

I'm reopening this issue as there are still a number of error types in the public API which are not public themselves, notably:

• ConnectionError
• AppDescriptorError
• RomError
• MissingPartition
• MissingPartitionTable

The partition table errors are especially problematic, as esp-idf-part is not stable (and likely won't ever be).

[MabezDev]

Oops, I think some of those were added after the initial PR by me 😅. Do you know if there is a lint or check we can do to prevent PRs being merged like this?

[jessebraham]

I was not able to find one unfortunately, perhaps I just missed it though. Seems like something that should exist to me.

[bugadani]

somewhat related rustc issue