JSON Reporting

[fosterbrereton]

This PR adds a new flag to ORC's config file, output_file_mode. Fixes #79.

The flag determines what format should be used for content written to external report files defined by either output_file or relative_output_file. The options are text or json. text will mirror all output sent to stdout and stderr to the file. json will limit output file contents to the ODRV report. This mode enumerates all object files with conflicting symbols, so can be quite verbose. It is intended for programmatic consumption. The default value is text.

This PR also removes the resource_metrics capability of the tool. It's been decided Tracy is a better solution to meet such a need. Therefore the flag and the code associated with supporting it have been pulled out.

An example snippet of the JSON output looks something like:

{
  "synopsis": {
    "dies_processed": 163925,
    "dies_skipped": 122239,
    "dies_skipped_pct": 74.57007777947231,
    "object_files_scanned": 92,
    "unique_symbols": 3435,
    "violations": 11
  },
  "violations": {
    "example_struct": {
      "attributes": [
        "structure:byte_size",
        "structure:calling_convention"
      ],
      "definitions": [
        {
          "attributes": {
            "byte_size": 624
          },
          "count": 33,
          "locations": {
            "/path/to/library/header.h:430": {
              "/path/to/library.a": [
                "foo.o",
                "bar.o",
                "baz.o",
...

[fosterbrereton]

This is where the JSON object is serialized to the file.

[fosterbrereton]

This circumvents writing anything to the file if we're in json output mode.

[leethomason]

Looks good!