[storage] fix: used the typed_store error type

[huitseeker]

This demonstrates the use of the following PR:
MystenLabs/mysten-infra#6

Please go and review the mysten-infra PR, once it's merged I'll update dependency pointers here.

This updates the typed_store version to one with a TypedStore error type, rather than using eyre::Error.
As a consequence, it allows a more fine-grained conversion of store errors to FastPay::StorageError.

[gdanezis]

Good to go! Many thanks for doing this clean-up after the nice talk you gave on the subject!

[gdanezis]

Weird question, I know: this may lead us to return to the client errors straight from the DB including detailed errors about encoding / decoding. Is there any case where this might leak security sensitive information through this error reporting?

[huitseeker]

The question is perfectly valid for the whole FastPayError hierarchy. One of the valid things to do here is to make sure your e.g. crypto signing lib doesn't share TMI, which is one of the things we should add as a TODO in #101 or similar. But there may be more scope of inspection here?

[gdanezis]

Is there no general Into() from Result<R,F> to Result<R',F'> if there is an R into R' and F into F'?

[huitseeker]

result1.map_or_else(|err| if err.is_nice() { Ok(foo(err)) } else { Err(bar(e)) }, |mkay| if mkay.is_cool() { Ok(baz(mkay)) } else { Err(quux(mkay)) }

But the idiom here is more probably :

let res = faillible()?;
Ok(res)

[oxade]

Very nice to see this in action
Will try to incorporate in client logic