Stop hardware write_memory from discarding a successful write on resume failure

[harshverma27]

Fixes #49

Summary

• HardwareBackend::write_memory chained halt → mww → resume behind a single ?-propagating async block. A resume failure after a successful mww write propagated as the whole function's error, discarding the fact the write itself had already landed — and (mirroring issue hardware read_memory: early ? skips resume, leaving STM32 permanently halted in OpenOCD #48) an mww failure would skip resume entirely, leaving the target halted.
• write_memory now tracks write_result and resume_result separately: resume is always attempted once halt has actually succeeded, and a resume failure is recorded via record_failure without overwriting a successful write result.

Test plan

• Added write_memory_keeps_a_successful_write_even_when_resume_fails: a mock telnet server that lets the mww write land, then forces the connection closed so the subsequent resume fails — asserts the function still returns Ok(()) and the failure is recorded.
• cargo test -p nucleus-hil --lib — 64 passed.
• cargo clippy -p nucleus-hil -p nucleus-cli --all-targets -- -D warnings — clean.
• cargo fmt --all -- --check — clean.

---

Generated by Claude Code