Batch annotation with RAM caching for slow disks

**Is your feature request related to a problem? Please describe.**
I'm annotating images and the remote machine which currently hosts my label studio environment has slow I/O therefore if I add or update an annotation I must wait the annotation to be written on disk before going to the next one the wait time is around 6 secs most of times longer then what takes me to annotate the image

**Describe the solution you'd like**
I'd like a batch annotation mode where I can select a range of tasks (e.g., by ID range or filter), which are then pre-fetched and cached client-side. I can then annotate all tasks in the batch without triggering any disk writes, and submit all annotations at once via a "Submit batch" button. Optionally, before fetching, the UI could display the estimated data size so users can confirm they have sufficient resources available.

**Describe alternatives you've considered**
I developed a custom annotation app tailored to my use case. It fetches tasks via Label Studio's API, lets me annotate images locally (loading them into RAM), and submits batches back via the API. I think this batch annotation workflow would be valuable as a built-in feature to help users with slow disks

**Additional context**