fix(nzbfs): bind file context to HTTP request context and reduce server WriteTimeout

[DarkseidAM]

Description

This PR addresses an issue where WebDAV reads can hang indefinitely when Usenet segment downloads stall or network connections drop silently.

Root Cause

1. Unbounded WebDAV server timeout: http.Server.WriteTimeout was hardcoded to 30 * time.Minute in setup.go. If a client (rclone) timed out and dropped the connection, Altmount kept the response goroutine open for up to 30 minutes.
2. Decoupled file context: OpenFile passed a context that did not automatically cancel when an HTTP request disconnected.

Changes

1. Request-bound File Context: Wrapped OpenFile context in a child context derived from the HTTP request context (fileCtx, cancel := context.WithCancel(ctx)), ensuring client disconnects automatically cancel in-flight UsenetReader downloads.
2. HTTP Server WriteTimeout: Reduced server WriteTimeout in setup.go from 30 minutes to 3 minutes to automatically close idle/stalled HTTP streams.

[javi11]

I think this can be achieve by seting up a context with cancel timeout on open the file, you don't need any watchdog

[DarkseidAM]

Refactored as suggested in the force-push: removed the per-file startIdleWatchdog goroutine entirely.

The PR now relies on:

1. context.WithCancel(r.Context()) bound to each OpenFile handle so HTTP client disconnects immediately propagate cancellation to in-flight NNTP segment downloads.
2. Reduced HTTP server WriteTimeout (3 minutes) in setup.go to ensure idle/stalled HTTP response streams auto-close and cancel their request context.

[javi11]

Now you add this cancelCtx but you don't use it anywhere

[DarkseidAM]

Apologies, forgot to stage the commit
Updated MetadataVirtualFile.Close() to invoke mvf.cancelCtx() when the file handle is closed.