GC infra improvements

[ghost]

• Remove name dotnet-gc-infra from the README is this is not named that any more.
• Remove FirstToLastEventSeconds metric, prefer TotalSecondsTaken instead.
• host_info.yaml now describes the processor ranges contained by each numa node.
• Support finalizable objects in GCPerfSim, similar to pinned objects
• Add --show-reasons flag to analyze-single
  • print a summary of reason counts over all GCs
  • print reasons for each individual GC in the single-gcs section
• Enhance --print-processes with support for arbitrary run metrics
  • Will show nothing if not a CLR process
  • To do this, must support getting ProcessedTrace objects for each individual process from a trace,
    resulting in some refactoring in process_trace.py
• Added --no-readjust parameter to make_memory_load.c.
  This means there is only an initial memory load but it never frees or allocates afterwards.
  This may be more fair as a test, as a GC that consumes less memory shouldn't be punished by another process taking up more memory.
• Support coreclr_specific in a config. This allows configs to be different for different coreclrs.
  This was useful as older coreclrs parsed complus_gcheapaffinitizeranges differently.
• Convert test_status metrics to direct properties of ProcessedTrace so they're easier to use from jupyter notebook
• Add new metrics AllocedMBAccumulated and MemoryPressure
• Add new configurations, and bgc_stepping, bgc_tuning_loh, bgc_tuning_soh reasons.
  (For testing BGC tuning using the free list with PI loops coreclr#26695)
• Support all document output args (e.g. --output-width) on all document-creating commands
• Add NO_DEFAULT as an alternative to MISSING; allows a dataclass to inherit from another which has fields without defaults.
• pylint bug import-outside-toplevel does not check from x import y imports pylint-dev/pylint#3175 is fixed, now must add more disables where this was violated