Add output validation to benchmarks for regression testing (Issue #267)

[Copilot]

Regression tests currently only verify that benchmark functions execute without errors — they don't validate correctness of the output. This adds a validate_output hook to each benchmark that the regression framework calls after each successful invocation.

Interface

• Added optional validate_output(input_config, output) -> bool to BenchmarkModuleInterface (defaults to True for backwards compatibility)
• Added Benchmark.validate_output() method that delegates to the benchmark's input.py if the function is defined
• regression.py now calls benchmark.validate_output(input_config, ret.output.get("result", {})) after each successful trigger invocation; failures are logged and mark the test as failed

The output passed to validators is the function handler's full return value (containing result and measurement keys).

Per-benchmark validators added to input.py

Benchmark	Validation logic
010.sleep	result == input['sleep'] (exact match)
110.dynamic-html	Non-empty HTML string containing the username
120.uploader	Non-empty storage key, URL echoed back correctly
130.crud-api	GET returns expected fields; PUT returns {}
210.thumbnailer, 220.video-processing, 504.dna-visualisation	Non-empty storage key in result
311.compression	Result key ends with .zip
411.image-recognition	Non-empty class label string and non-negative integer index
501.graph-pagerank	Float in [0.0, 1.0]
502.graph-mst	Result is not None
503.graph-bfs	Non-empty list/tuple

Example

# benchmarks/000.microbenchmarks/010.sleep/input.py
def validate_output(input_config: dict, output: dict) -> bool:
    return output.get('result') == input_config.get('sleep')