fix: decode uppercase segwit addresses

[modelsbridgeaicom-ship-it]

Summary

• Normalize Bech32/SegWit HRPs when decoding addresses so all-uppercase BGL1... addresses decode like their lowercase form.
• Reuse the same parsing in address_to_hash, address_to_script, address_type, address_net_type, and get_witness_version.
• Add focused regression coverage for uppercase mainnet SegWit decoding.

Why

Bech32 addresses are valid when entirely lowercase or entirely uppercase. Before this change, the decoding helpers only recognized lowercase bgl / tbgl prefixes, so an uppercase address could fail hash/script conversion even though it represents the same address.

Validation

• python -m compileall -q pybgl tests.py setup.py setup_tools.py
• Focused RED/GREEN regression using local lightweight native-extension stubs:
  • origin/master returned None / UNKNOWN for uppercase Bech32 decode paths.
  • this branch returns the same hash/script/type/network/witness version as lowercase.
• git diff --check passes with Windows line-ending warnings only.

Bounty context: submitted for the Bitgesell PR bounty program: BitgesellOfficial/bitgesell#39. Payment details can be provided if maintainers accept the PR and request them.

[MyTH-zyxeon]

Maintainer-facing review assist for the Bitgesell #39 bounty lane:

This PR is a focused bug fix for uppercase Bech32/SegWit address decoding in pybgl/functions/address.py plus regression coverage in pybgl/test/bech32_address_decode.py (+51/-14). It is currently OPEN/MERGEABLE/CLEAN with no prior PR comments or review threads visible.

What looks good:

• The helper centralizes SegWit HRP splitting and normalizes BGL/TBGL prefixes before the decode helpers branch.
• The touched public helpers are the right surface for this failure class: address_to_hash, address_to_script, address_type, address_net_type, and get_witness_version.
• The new regression checks uppercase mainnet P2WPKH equivalence against the lowercase address across hash, script, type, network, and witness version.

Review points before approval:

• Confirm whether rebase_32_to_5() intentionally rejects mixed-case Bech32 payloads. Bech32 permits all-lowercase or all-uppercase strings, but mixed-case addresses should stay invalid; if the decoder accepts both alphabets in one payload, add a mixed-case rejection fixture.
• Consider adding one TBGL1... uppercase testnet fixture so the TESTNET_SEGWIT_ADDRESS_PREFIX path is covered too.
• If maintainers require full package confidence, ask for a plain python tests.py or equivalent package-level run in addition to the focused compile/RED-GREEN evidence already listed.

If those checks line up, this looks like a real Bitgesell-owned bug fix rather than a metadata-only cleanup, so it is a reasonable #39 bounty-review candidate.