Condense binary cataloger config in JSON output

[wagoodman]

The last rebase of #1383 resulted in bringing in changes from #2469 , which makes the configuration in the JSON output prohibitively large:

syft <anything> -o json | jq '.descriptor.configuration.packages'

    "packages": {
      "binary": {
        "classifiers": [
          {
            "class": "python-binary",
            "cpes": [
              {
                "Edition": "",
                "Language": "",
                "Other": "",
                "Part": "a",
                "Product": "python",
                "SWEdition": "",
                "TargetHW": "",
                "TargetSW": "",
                "Update": "",
                "Vendor": "python_software_foundation",
                "Version": ""
              },
              {
                "Edition": "",
                "Language": "",
                "Other": "",
                "Part": "a",
                "Product": "python",
                "SWEdition": "",
                "TargetHW": "",
                "TargetSW": "",
                "Update": "",
                "Vendor": "python",
                "Version": ""
              }
            ],
            "fileGlob": "**/python*",
            "package": "python",
            "purl": {
              "Name": "python",
              "Namespace": "",
              "Qualifiers": [],
              "Subpath": "",
              "Type": "generic",
              "Version": "version"
            }
          },

...

Note that the binary classifiers are now shown, however, every field in CPEs is shown (as well as others on the config).

This PR makes a few changes:

• Marshalling the binary Classifier struct will serialize CPEs to a string
• Marshalling the binary CatalogerConfig struct will only show the class names configured. Since the behavior is not reproducible from a description alone (the evidence parameter, the most important one, is a function) it is not really worth showing the additional fields.