refactor: 🔒 Reduce file storage locks critical zone

[HermanObst]

Problem

File storage locks were being held for longer than necessary:

1. file_download_manager.rs: Write lock acquired before try_download_chunk_batch and held during network I/O, retries, and 500ms delays between attempts.

2. msp_upload_file.rs: Write lock acquired early in handle_new_storage_request_event and held across file existence checks, even when only a read lock was needed.

Solution

file_download_manager.rs:

• Moved lock acquisition into process_proven_chunk
• Lock now acquired only for the brief duration of writing each chunk
• Network I/O happens completely outside any file storage lock

msp_upload_file.rs:

• Changed file_in_file_storage check to use a scoped read lock instead of holding the write lock
• Write lock now only acquired when actually inserting a file
• is_file_complete check also uses a scoped read lock

---

TBD: It can be argued that taking and releasing the write lock for every chunk would be a lot of overhead:

• RwLock are pretty fast, even more compared to disk writes (what we are already doing by chunk)
• To be seen if this decrease performance

Anyway, open to discussion :)

[ffarall]

Great job!