feature: mpl-utils: Log assert failures for easier debugging

[KartikSoneji]

At the moment, when an assertion fails, the program only throws an error about the type of assertion that has failed.
This there no indication to which account is the incorrect one, making it a nightmare to debug.

For example:

> Program invoked: Token Metadata Program
  > Program logged: "IX: Delegate"
  > Program logged: "Public key does not match expected value"
  > Program consumed: 18426 of 373338 compute units
  > Program returned error: "custom program error: 0x9a"

This PR adds logs that print the account that failed the assertion and some debug information about why the assertion failed.
They only trigger if the assertion fails, so shouldn't effect the compute unit budget during normal operation.

[blockiosaurus]

Thanks for this PR! It is indeed awful to debug any of these issues. The lack of logging of specific failures is unfortunately a necessity due to program size and compute constraints.

However, ExtendProgram is scheduled to be enabled Monday so that will hopefully solve the size issue.

[KartikSoneji]

Oh does calling sol_log() really increase the program size so much?
Or is it format!()? I can refactor to remove calls to format if that's the issue.

[blockiosaurus]

It's the actual strings. I think we have more space now but we were a handful of bytes from the max a couple months ago.

[KartikSoneji]

Ah I see.
Let me know if the messages are too long, I think I can cut it down to just a few extra bytes.
Eg: <actual key>!=<expected key>
Even terse log messages like this should be a great help while debugging.

[KartikSoneji]

Hi, just a friendly ping on this.

[blockiosaurus]

Huzzah, we have space now! Can you just change these to use msg! to be consistent with our other programs? That should also be a little cleaner as it uses the same string formatting as println!.

[KartikSoneji]

Done!

[febo]

"invalid owner for {key} ({actual} != {expected})"

[febo]

"invalid derivation {actual_key} != {key}"

[KartikSoneji]

Looks like \n is not needed, I'll remove that.

[febo]

What do you think if we make the log messages a bit shorter so the log does not get cluttered?

[KartikSoneji]

Hmm I really think the log messages should provide as much information to debug as possible.
Plus I think in most cases programs failing an assertion is a fatal error, so this should be the last message that is in the log.

[febo]

Hmm I really think the log messages should provide as much information to debug as possible. Plus I think in most cases programs failing an assertion is a fatal error, so this should be the last message that is in the log.

I agree that log messages should be descriptive, but the observation here is because this is a library not a program. A program might handle an assert failure so it does not end up being a "fatal error".

[KartikSoneji]

OK, I'll shorten the messages for assert_signer, assert_initialized and assert_rent_exempt.
But I think assert_owned_by, assert_derivation are almost always fatal errors so should clearly communicate the issue without having to refer back to the source code of the program.