Disable Automatic Algorithm Fallback

[jonathanknowles]

Related Issue

#30

Summary

The cardano-wallet library currently performs automatic fallback from Random-Improve to Largest-First if the former algorithm is unable to compute a valid coin selection.

However, this automatic fallback might not be appropriate for all users of the coin selection library.

This PR:

• removes the automatic fallback logic within the implementation of Random-Improve.
• adds an appropriate error handling function, since Random-Improve can no longer rely on the error handling provided by Largest-First.

The error handling is already tested by our existing test suite.

[paweljakubas]

if we disable largest first fallback, would be good to show in testing that we can recreate old setup. by invoking random and then rely on largestFirst. And here deliver test showing that we do not have fallback (there were test like that to my best knowledge in cardano-wallet), and when having the composition with fallback it service those cases.

[jonathanknowles]

@paweljakubas wrote:

if we disable largest first fallback, would be good to show in testing that we can recreate old setup. by invoking random and then rely on largestFirst. And here deliver test showing that we do not have fallback (there were test like that to my best knowledge in cardano-wallet), and when having the composition with fallback it service those cases.

Thanks for raising this point!

I agree, it's very important to preserve the existing behaviour of cardano-wallet when we eventually reuse cardano-coin-selection from cardano-wallet, replacing the old code.

I've added an item to the acceptance criteria for this task here, which highlights the need for testing to make sure that fallback continues to happen as expected.