Replace panics and simple errors with custom error types

[ivanlele]

One of the PRs in a series addressing issue #28.

In the legacy commands (address, block, tx), panic-based logic has been removed from the execution path and replaced with custom errors.

For the remaining commands, crate::cmd::simplicity::Error has been removed from the execution logic in favor of command-specific custom errors.

This change is crucial for enabling programmable interfaces, as it allows errors to be caught and handled programmatically by inspecting the underlying error chain rather than relying on catching panics or parsing of string errors.

[ivanlele]

CI passed: 8384b9

[apoelstra]

In a548ad8:

Rather than doing let inspect_fn = || -> Result can you move the logic into exec_inner functions the way that it's done in the new simplicity commands?

[apoelstra]

In a548ad8:

Heh, this is unfortunate. It's ultimately a bug in rust-bitcoin's ParsePublicKeyError that its errors have a terrible Display impl. Nothing we can do here.

[apoelstra]

On 8384b92 successfully ran local tests

[ivanlele]

In a548ad8:

Rather than doing let inspect_fn = || -> Result can you move the logic into exec_inner functions the way that it's done in the new simplicity commands?

In the Err match arm for output handling, I kept panics to trigger the existing panic hook. This preserve the same output formatting and as their old counterparts, even though the newer commands use cmd::print_output

[apoelstra]

Can you squash df35c9c into a548ad8 so that there isn't one commit which rewrites the diff of another?

[ivanlele]

Can you squash df35c9c into a548ad8 so that there isn't one commit which rewrites the diff of another?

done

[apoelstra]

ACK c3b23ec; successfully ran local tests; definitely an improvement, though we will want to go through and break up these errors later to avoid massive enums where most variants are not produced from most functions