fix: decode block header fields from Core JSON

[tolga-tom-nook]

Summary

• Decode Bitgesell Core getblockheader bits as the compact hex string returned by the RPC API.
• Fix previousblockhash / nextblockhash JSON tags so lower-case Core response keys populate the struct.
• Add an httptest regression covering a Core-shaped getblockheader response.
• Harden the regression test's server URL parsing with net/url instead of brittle string splitting.

Compatibility note

BlockHeader.Bits intentionally changes from uint32 to string. Bitgesell/Bitcoin Core RPC returns bits as a compact target hex string such as "1d00ffff", and this also matches the existing Block.Bits string shape in this package.

Validation

• RED on upstream shape with the new regression: go test ./... fails before the fix because BlockHeader.Bits is uint32 while Core returns "1d00ffff".
• GREEN after fix: go test ./... passes.
• Revalidated after test hardening: go version => go1.22.2 linux/amd64; go test ./... passes; git diff --check clean.

Bounty context

Submitted for the Bitgesell improvement PR bounty:

• BGL (Bitgesell) Bounty/Improvement Program ($50000 budget) bitgesell#81
• BGL PR bounty hunt ($10000 overall budget) bitgesell#39

Payout rail if approved: USDT/EVM-compatible 0x4a76c7E64C08cF29B59eFC640b4ada97A270d428.

[MyTH-zyxeon]

Maintainers: diff-only review-assist for Bitgesell improvement lane #81 / #39.

This PR looks narrow and reviewer-friendly from the public diff:

• bgld.go aligns BlockHeader.Bits with the Core RPC response shape by decoding the compact target as a string.
• The JSON tags for previousblockhash / nextblockhash are corrected to the lower-case field names Bitgesell Core actually returns.
• blockheader_json_test.go adds a focused httptest regression using a Core-shaped payload, so the acceptance path is visible without needing a live node.

Suggested acceptance checks:

• Confirm downstream callers are not depending on BlockHeader.Bits being uint32; the change looks correct for RPC compatibility, but it is a public type-shape change.
• Keep the new regression as the canonical proof that getblockheader matches Core's wire format (bits string plus previous/next hash keys).
• If maintainers want extra confidence, one more lightweight assertion on omitted previousblockhash / nextblockhash for tip blocks would close the last obvious edge without broadening scope.

From a bounty-review perspective, this seems to fit the issue's "small, concrete API/test correctness fix" lane better than a docs-only cleanup. If merged, it would help to route payout review back through BitgesellOfficial/bitgesell#81 so contributors are not split across stale repo threads.