simc: error: refactor generic String to error types

[gerau]

For motivation see #324 (comment).

[KyrylR]

Is it expected that let x: u8 = 999 now reports Integer parsing error, losing the previous number too large to fit in target type detail?

[gerau]

I've based my changes on this comment: #324 (comment), and if IIUC, this is pretty much expected. I would be happy to change it if you want.

[KyrylR]

ACK 1448898; successfully ran local tests

[apoelstra]

Is it expected that let x: u8 = 999 now reports Integer parsing error, losing the previous number too large to fit in target type detail?

Our reports should use Error::source to provide the full stack of error details, similar to how anyhow outputs errors. The other method, of manually pasting strings together in the errors' Display impls so that you can one-shot fmt the whole error stack, is (at best) inflexible since it prevents us reformatting stuff, and at worst loses information because 3rd-party errors will expect us to call source to get inner errors.

I also don't think "everything in fmt" it will survive the transition to multi-line errors.

[apoelstra]

In 6a5099d:

You need to impl the source method for Error.

[apoelstra]

Oh, never mind, this is done in 1448898

[apoelstra]

ACK 1448898; successfully ran local tests