Track Bitcoin target in proxy channel from SetNewPrevHash nbits

[rx18-eng]

Body

fixes the no-tp side of dmnd-pool/dmnd-client#111

so right now if youre running translator-style proxy with no TP connected, the non-jd branches in channel_factory just hardcode bitcoin_target = [0; 32]. which means OnNewShare::ShareMeetBitcoinTarget literally cant fire for us coz no hash is ever <= 0. so a share that actually finds a block looks like a normal share, hits the 70/min rate limiter on the dmnd-client side, and if were unlucky it just gets dropped. real block lost.

went with what @Fi3 suggested in the thread - just read nbits from the SetNewPrevHash the pool already sends us. we trust the pool for jobs+prev_hash anyway so its not a new trust assumption, no new deps, just a tiny nbits_to_target helper that wraps bitcoin::Target::from_compact and gives back a mining_sv2::Target.

actual diff is small btw - bits was already being extracted right under the [0;32] line and passed to check_target. i just moved the bitcoin_target line down a few lines so it can use bits, and dropped the .into() at the call site since the helper returns Target directly.

fail-open semantics preserved (the constraint @Fi3 flagged - wrong target must never block a share):

• nbits=0 → Target::ZERO → no hash matches → share takes the normal path. nothing dropped.
• super lax nbits like regtest → max target → some shares get classified as block-finds and go upstream un-gated. pool just sees em as normal low-hash shares. still nothing dropped.

so worst case wrong target = wrong classification, never a dropped share.

Tests i added (3 unit tests on the helper):

• nbits_to_target_zero_yields_zero_target so malformed pool msgs cant produce false-positive blocks
• nbits_to_target_matches_existing_test_helper cross-checks against the nbit_to_target fixture thats already in this same test module
• nbits_to_target_regtest_max_is_greater_than_mainnet quick sanity check on Target ordering since check_target does hash <= bitcoin_target

the actual integration (the ShareMeetBitcoinTarget arm becoming reachable end to end) ill cover in the dmnd-client bridge PR - will link it back here once this merges. existing JD-path test test_complete_mining_round still passes so didn't break that side.

cc @jbesraa @Priceless-P @Fi3

[rx18-eng]

@Fi3 @Priceless-P would appreciate your thoughts on this !