Feat/deployment preparations

[jomkv]

Added

• Docker setup for deployment at VM
• Adjusted threads and concurrency to handle heavy load
• Optimized remarks generation by removing BART completely
• Excluded model training endpoint on prod as it is used during development only

TODO

• Optimize resource usage, VM's CPU is always under heavy load during concurrent requests (which is expected). If we can find a way to shift more of the workload to the memory that would be great
• Investigate why accuracy scores are lower when using deployed DB

I'll test this branch on the VM tomorrow to see if anything broke, just wanted to create a PR ASAP so the accuracy test can run on github action while I sleep 😃

[github-actions]

Accuracy Check Results

Accuracy evaluation did not produce summary JSON in step output.

• Evaluation step outcome: failure/success
• Run: https://github.com/jomkv/truescope-api/actions/runs/24820517556

[Mark-0909]

Goods for me