Move NVBench JSON result parsing into cuda.bench.results with explicit
BenchmarkResult, BenchmarkResultDevice, BenchmarkResultSummary,
SubBenchmarkResult, and SubBenchmarkState types. Remove the result reader
from the top-level cuda.bench namespace and require construction through
BenchmarkResult.from_json() or BenchmarkResult.empty().
Preserve bulk sample/frequency parsing and estimator helpers while making
summaries rich objects that retain tag/name/hint/hide/description metadata.
Add nvbench-json-summary to render NVBench JSON output as an NVBench-style
markdown summary table, including axis formatting, device sections, hidden
summary filtering, and summary hint formatting.
Update packaging, type stubs, and tests for the new namespace, renamed
classes, Python 3.10-compatible annotations, and summary-table generation.
