Allow additional custom tests in the generator

[junedev]

For some exercises it makes sense to add some language specific test cases additional to what is listed in the problem spec test cases.

Here an example where this would be helpful: #1451

Let's discuss ideas how to adjust the generator to incorporate additional test cases. The restriction should be that the new cases adhere to the format of the existing ones (e.g. no new properties are introduced).

[andrerfcsantos]

What about having a canonical-data.append.json in the .meta directory for the exercise?

[junedev]

@andrerfcsantos Great idea! That shouldn't be hard and the generator is in that directory already anyway.

So the task here is to change the generator so that it checks whether .meta/canonical-data.append.json is present for the exercise. If it it is, the content should be combined with the content of the usual canonical-data.json file.

[eklatzer]

I will have a look when doing #2279

[junedev]

@eklatzer Not sure whether you mean you want to include this in #2279 but if yes, it would be better if this would be separate PR. Otherwise if the added code for this feature has issues that blocks the other PR for now reason. And the more things are in one PR, the harder it is to review.

[eklatzer]

I will wait for #2279 to be merged and then I will do it

[eklatzer]

Note for future: we could use the feature of this issue in the exercise reverse-string as there is currently the following additional test case:

// mutiByteCases adds UTF-8 multi-byte case,
// since the canonical-data.json (generator data source for cases_test.go)
// doesn't have any such cases.
var multiByteCases = []reverseTestCase{
	{
		description: "a multi-byte test case",
		input:       "Hello, 世界",
		expected:    "界世 ,olleH",
	},
}