native-token: Deprecate float usage, add str usage

[joncinque]

Problem

The sol_to_lamports and lamports_to_sol functions are problematic because they use floats, which are not precise.

Summary of changes

Deprecate the functions, and add versions that use strings instead.

Added @tao-stones to check the FeeStructure change -- it looks like FeeStructure::new is only used in one test.

[rustopian]

Correct, good tests, all works great.

[rustopian]

Oops, @tao-stones your review is still requested directly, my bad on triggering auto-merge

[t-nelson]

i do this far too often 🤦

[joncinque]

Ah darn, sorry -- for the first point, we'll get a warning on removal, which will fail CI, so we can't forget to remove the const.

And for the second point, I wasn't aware of that function... I guess the options are to either include a bool flag (edit: in this new function), or just remove it and tell people to use build_balance_message. All of the users of lamports_to_sol are CLIs / user-facing, so dependencies aren't an issue. I'd lean towards the second option. What do you think?

[tao-stones]

Changes to fee-structure is good.

[t-nelson]

Ah darn, sorry

no worries! i'm the idiot that didn't hit the comment button

for the first point, we'll get a warning on removal, which will fail CI, so we can't forget to remove the const.

👍

And for the second point, I wasn't aware of that function... I guess the options are to either include a bool flag (edit: in this new function), or just remove it and tell people to use build_balance_message. All of the users of lamports_to_sol are CLIs / user-facing, so dependencies aren't an issue. I'd lean towards the second option. What do you think?

yeah agreed. i think we should stick one implementation to avoid all of the caveats of replicode.

... though i could be convinced that build balance message belongs in a different crate 🤔

[joncinque]

Well, I put in #174 so we can discuss there