Performance regression test cases

[MichaelBelousov]

We should probably run each case sequentially on the current iModel from the battery of iModels to test:

const testCases = {
  "identity transform": require("./identity-transform"); 
};

// need to not download iModels that we don't intend to test for local experiments 
for (const iModel of iModelsInProject({ filter: m => m.tShirtsize === 'm' } )) {
  describe(`Transforms of ${iModel.name}`, () => {
    for (const [testCaseName, testCaseRun] of Object.entries(testCases) {
      it(testCaseFunction.name, function () {
        //if (!process.env.CI && iModel.tShirtsize !== 'm') this.skip();
        testCaseRun(iModel);
      });
    }
  });
}

Sorted by priority:

• [x] identity transformation (source to empty target)
• [x] prepare fork
  • [x] prepare fork all fed guids
  • [x] prepare fork few fed guids
  • create elements with "empty" fed-guid (000-0000...), @itwin/core-bentley should have something for that
  • [x] prepare fork few fed guids but add them to the source ourselves
  • copy the source in the test, and use the withSqliteStatement API to query the bis_Element.FederationGuid and update it to have a federation guid if it doesn't already (use randomblob(16))
  • [x] raw no transformer
  • we can do this one later. Insert target scope element, iterate over the elements and add ExternalSourceAspects to those without fedguids, as the transformer would, do not invoke the transformer.
• [ ] synchronize fork (many/few fed guids)
• [ ] reverse synchronize fork (many/few fed guids)
• [ ] physical model consolidation
• [ ] many deletes
• [ ] many relationships
• [ ] complex schema
• [ ] complex geometry
• [x] control
  • [x] raw insert many elements
  • [x] apply changeset containing same elements
  • [x] populate empty target from source containing same elements
  • [x] populate empty target from source containing same elements, also add provenance

[MichaelBelousov]

closing as close enough