Add support for additional datasets beyond SWE-Bench Lite

[kabirgh]

SWE-bench_Lite is hardcoded as the only dataset that Agentless can run on.
I'd like to evaluate performance on datasets like SWE-bench_Verified or other private datasets.