chore(build): optimize cargo dev profile and prune crate-type to reduce storage footprint

[huwany1]

Summary

Reduces local development storage overhead and prevents runaway build cache accumulation:

• Prunes crate-type in src-tauri/Cargo.toml to ["rlib"] for desktop builds, eliminating the redundant monolithic staticlib archive (~4 GB single file on Windows MSVC) and temporary .a outputs.
• Adds [profile.dev.package."*"] debug = 1 to generate line-level debuginfo for third-party dependencies while preserving full debug = 2 for the main crate so rich IDE variable inspection remains intact.
• Adds clean:target convenience script to package.json.

Behavioral Impact

• Build cache size in src-tauri/target is reduced by tens of gigabytes across iterative runs without degrading local debugging capabilities.
• Third-party dependency PDBs and intermediate objects shrink significantly.
• Clean build verified and src-tauri/target drops to ~3.9 GB.

Verification

• cargo check --manifest-path src-tauri/Cargo.toml (Passed in 46.8s)
• npm run check:ipc (Passed: 171 registered commands)
• npm test (Passed: 69 test files, 617 tests)

[chyinan]

审查完成，整体 LGTM，没有发现需要阻塞合并的功能、安全或事务一致性问题。GitHub CI 6/6 通过；本地精确提交上 characters::activation_tests 35 项、commands::characters 39 项通过。非阻塞建议：建议在 src-tauri/src/characters/activation.rs:431 附近补一个 max_connections(1) + apply/restore failure 的回归测试，覆盖事务在 recover_committed_backend 前释放连接的死锁场景。除此之外可以合并。