Fix how we fetch job logs

[golharam]

get_job_log wasn't returned AWS CloudWatch logs. There were a few issues identified by Claude:

Performance Issues

1. Synchronous Blocking I/O
   The get_log_events() function makes synchronous boto3 calls to CloudWatch Logs, which blocks the entire request thread while waiting for AWS responses.

2. Fetching All Logs at Once
   Lines 219-234 in services.py fetch ALL log events in a while loop until exhausted. For large jobs with millions of log lines, this causes:

Memory accumulation (all logs stored in events list)
Extended processing time (multiple AWS API calls in sequence)
Gateway timeout before response can be sent

3. Inefficient boto3 Client Creation
   Line 221-224 creates a new boto3 client inside the while loop on every iteration, adding unnecessary overhead.

4. No Pagination Support
   The API endpoint doesn't support pagination - clients must wait for the entire log file, even if they only need recent entries.

5. No Caching
   Logs for completed jobs never change, but they're fetched from CloudWatch every time.

@EricSDavis - I ended up implementing a StreamingResponse which seems to be the best solution here, however the FrontEnd still isn't displaying anything.

[golharam]

Closing in favor of #215