Control plane support for Extractor regex replace

[ben-taussig-solo]

Description

• This PR adds control-plane support for the extractor regex replace functionality introduced in Add regex replace functionality to transformation filter extractors solo-io/envoy-gloo#301
• That PR extended the transformation filter Extractor functionality to support regex replace/replace all functionality.
• This PR allows Gloo Edge users to configure this new extractor functionality
• This PR also adds unit and e2e tests against the new functionality to ensure that it works as expected.

User-facing API

• Extractors now support a mode field, which represents the type of extraction to perform. The mode field can be one of EXTRACT, SINGLE_REPLACE, or REPLACE_ALL. The mode field is optional, and defaults to EXTRACT (prior to these changes, EXTRACT was the standard behavior for extractors).

  // The mode of operation for the extraction.
  enum Mode {
    // Default mode. Extract the value of the subgroup-th capturing group.
    EXTRACT = 0; 
    // Replace the value of the subgroup-th capturing group with the replacement_text.
    // Note: replacement_text must be set for this mode.
    SINGLE_REPLACE = 1;
    // Replace all matches of the regex in the source with the replacement_text.
    // Note: replacement_text must be set for this mode.
    // Note: subgroup is ignored for this mode. configuration will fail if subgroup is set.
    // Note: restrictions on the regex are different for this mode. See the regex field for more details.
    REPLACE_ALL = 2;
  }

• Extractors now have a replacement_text field, which is used in SINGLE_REPLACE and REPLACE_ALL modes. This field is optional, and is required if the extractor is configured in SINGLE_REPLACE or REPLACE_ALL mode.

  // The string to replace the matched portion of the source with.
  // Used in SINGLE_REPLACE and REPLACE_ALL modes.
  google.protobuf.StringValue replacement_text = 5;

Control-plane Updates

• The control-plane now handles the new mode extractor field,
• The control-plane now has configuration validation behavior unique to each mode
  • If the mode is EXTRACT, the control-plane will error if replacement_text is set.
  • If the mode is SINGLE_REPLACE, the control-plane will error if replacement_text is not set.
  • If the mode is REPLACE_ALL, the control-plane will error if subgroup is set, and if replacement_text is not set.

Testing

• This PR adds unit tests for the new functionality, ensuring that the control-plane correctly handles the new mode and replacement_text fields, and that invalid configurations are rejected.
• This PR also adds e2e tests for the new functionality, ensuring that the new functionality interacts with the data plane as expected.

Additional Bug Fixes

• Additionally, this PR adds control-plane validation of regex/subgroup configuration, mirroring validation done in the envoy-gloo transformation filter
  • The check in the transformation filter evaluates whether the specified subgroup is larger than the total number of capturing groups in the regex pattern. If so, the filter throws an exception, preventing envoy from handling any traffic altogether until the configuration is fixed.
  • This PR duplicates that check in the control-plane, preventing this misconfiguration from being pushed to the envoy proxy in the first place.
  • However, the check requires the regex pattern to be compiled, which will hurt translation performance, especially if many regexes are being validated. If you have qualms about this, I am happy to move these changes into a separate PR. where we can discuss the tradeoffs in more detail.

Remaining Work

• I need to make substantial changes to the documentation before this work is complete. I may choose to handle that in a separate PR if that seems more straightforward
  • Update: these changes have been added
• I am going to leave this PR in WIP/draft state until the corresponding envoy-gloo PR is merged and released: Add regex replace functionality to transformation filter extractors [Revised] solo-io/envoy-gloo#309
  • Update: The envoy-gloo PR has been merged and released and I have marked this PR ready for review