experiment: make LLM_FIX_LIMIT adjustable (#316)

During development it's convenient to being able to adjust this, for
example when focus during development is on aspects of OSS-Fuzz-gen that
is unrelated to code fixing it's convenient to decrease this to speed up
evaluation.

experiment/evaluator.py

@@ -30,7 +30,7 @@
 from experiment.workdir import WorkDirs
 from llm_toolkit import code_fixer
 
-LLM_FIX_LIMIT = 5
+LLM_FIX_LIMIT = int(os.getenv('LLM_FIX_LIMIT', '5'))
 
 OSS_FUZZ_COVERAGE_BUCKET = 'oss-fuzz-coverage'