Upgrade bech32 dependency (iterative)

[optout21]

Continuation of #3244 . Fixes #3176 .

Changes summary:

• Upgraded bech32 dependency to 0.11.0 (from 0.9)
• The previously used type bech32::u5 can be replaced by bech32::Fe32, mostly as a drop-in replacement.
• u5::try_from_u8 is now Fe32::try_from
• bech32::Error is mostly replaced by bech32::primitives::decode::CheckedHrpstringError
• ToBase32 (and Base32Len) traits are discontinued in bech32, they have been reimplemented in iterative style as Base32Iterable. The implementations (Bolt11InvoiceFeatures/Payment*/...) have been moved to lightning-invoice crate (as the trait is defined there)
• FromBase32 trait is discontinued in bech32, it has been taken over to lighning-invoice crate. The implementations have been moved to lightning-invoice crate (as the trait is defined there)

Here are some minor issues identified

• Hrp/checksum parsing has changed (uses new iterative API)
• Fe32 does not have Ord, so it had to be explicitly implemented for RawTaggedField
• Fe32 does not have Default, for which some workaround was needed. This will be added in next bech32 0.12 release (see Add Ord trait to Fe32 (derived) rust-bitcoin/rust-bech32#186)
• Conversions of bech32 error types are not always optimal
• Parsing of tagged values in an invoice stays own implementation (not supported by bech32 crate)

TODO:

• fix fuzz target
• fix fmt
• squash
• Incremental-mutants

[codecov]

Codecov Report

Attention: Patch coverage is 93.87755% with 27 lines in your changes missing coverage. Please review.

Project coverage is 89.64%. Comparing base (66fb520) to head (7da1b51).

Files with missing lines	Patch %	Lines
lightning-invoice/src/de.rs	86.99%	10 Missing and 6 partials ⚠️
lightning/src/offers/parse.rs	63.63%	2 Missing and 2 partials ⚠️
lightning-invoice/src/test_ser_de.rs	97.00%	0 Missing and 3 partials ⚠️
lightning-invoice/src/lib.rs	96.55%	2 Missing ⚠️
lightning-invoice/src/ser.rs	98.65%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3270      +/-   ##
==========================================
- Coverage   89.64%   89.64%   -0.01%     
==========================================
  Files         126      126              
  Lines      102676   102753      +77     
  Branches   102676   102753      +77     
==========================================
+ Hits        92045    92109      +64     
- Misses       7912     7932      +20     
+ Partials     2719     2712       -7     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[optout21]

CI: A few misses in iterative-mutants on the Bolt11InvoiceFeatures serialization method (unchanged). To see if checks can be added, or the method can be rewritten in a simpler form.

[optout21]

CI: All builds OK. Iterative mutants is satisfied now, but the build was interrupted (?) after the tests (verified the logs + locally).

[optout21]

Added a minor cleanup: hash_from_parts is called from a single place (from signable_hash(); in the other place in the SignedRawBolt11Invoice parser, it's called through signable_hash()), this way one version is sufficient.

[optout21]

A note regarding bech32->bytes conversion with padding:
This is needed currently when computing hash of an invoice (hash_from_parts), and, if we want to switch for Features using bech32 library (as mentioned in a review comments above), also in the Features decoding.

I've raised an issue to rust-bech32, to possible offer an option for bech32->bytes conversion where last bits are not dropped but padded, that would be the nice solution.

I also realized padding can be done in an iterator adapter (counts the elements and and the end optionally outputs padding), so this padding can be applied nicely in the chained iterators approach (as discusses above).

[optout21]

Awaiting (re)review, @TheBlueMatt . Thx.

[TheBlueMatt]

This is shaping up nicely. We took this opportunity to avoid allocations in the serialization pipeline (sadly with allocations to avoid GATs...) but there's also a ton of really useless allocations in the deserialization pipeline, some of which we can trivially remove. We certainly don't have to fix it all in one go, but it'd be nice to reduce them where its easy here.

In any case, please feel free to go ahead and rewrite the commit history to have a logical progression so its easy to review :)

[optout21]

Commits squashed

[TheBlueMatt]

We should get a second reviewer on this.