Add per-process and per-call statistics for time and size metrics

[izzet]

This pull request introduces enhancements to the statistics and metrics computation pipeline, focusing on clearer separation between per-process and per-call metrics, and improving the calculation and naming of derived statistics. The main changes include new utility functions for column renaming and per-call statistics derivation, updates to aggregation logic, and expanded metric calculations for both process-level and call-level views.

Statistics and column handling improvements:

• Added build_view_rename_map to standardize column names for view-level statistics, ensuring per-process and per-call metrics are clearly distinguished and fixing naming inconsistencies.
• Introduced derive_call_stats, which computes per-call mean and standard deviation from sum-of-squares columns and cleans up intermediary columns.
• Updated aggregation logic in _compute_view and related functions to handle new per-call and per-process metric suffixes, ensuring correct aggregation functions are applied for each metric type. [1] [2] [3]

Metric calculation enhancements:

• Refactored metric suffixes and calculations in set_view_metrics to separately compute process-level (*_proc_*) and call-level (*_call_*) metrics, including new fractional metrics for both categories. [1] [2]
• Expanded the calculation of fractional and percentile metrics to cover both process-based and call-based statistics, improving the granularity of performance analysis.

Consistency and naming updates:

• Updated references and logic throughout the analyzer to use the new process-level metric suffixes (*_proc_*), replacing previous ambiguous suffixes and ensuring consistent metric naming in all views and additional metric calculations. [1] [2]

These changes collectively improve the clarity, accuracy, and flexibility of metric computation in the analysis pipeline.