Support proto3-suite 0.5.0 and later.

[j6carey]

Starting with proto3-suite 0.5.0, the initial character of
a service method names is no longer forced to uppercase.

[jcarr-awake]

This is clear enough for me, my I-dont-know-anything-about-the-original-change question is why we wouldn't be able to just use prefixedFieldName always, unless it didn't exist?

[Friede80]

I don't quite understand the difference between the two. The -ddump-splices output looks the same. (Although that very well could be the test file doesn't exercise this difference)

Also, we use prefixFieldName in several other places as well. Wouldn't they need to be updated as well?

[j6carey]

@Friede80, @jcarr-awake ,

prefixedMethodName was added by awakesecurity/proto3-suite#171 (first appearing in proto3-suite 0.5.0) and used for service method names only, with prefixedFieldName still used for other things. Apparently the desire is to make it possible for two methods of the same service to differ only in the case of the first letter of their method names.

I ran into a compilation failure in an Awake application of this library before making this change, because the Haskell spelling of at least one service method was wrong.

Regarding other uses of prefixedFieldName in grpc-mqtt: thanks for spotting that problem with this PR; I'll push up a change.

[j6carey]

c346d57 replaces the other uses of prefixedFieldName and isolates the conditional code to a compatibility module.

[Friede80]

This looks good. Thanks for moving it out to a separate module.

It does raise a separate issue that the tests fail when using proto3-suite >= 0.5.0, because the generated code is using version 0.4.3. I think we are still good to merge this PR and I'll create an issue for fixing the tests.

[j6carey]

@Friede80 , I think the test failures were a legitimate concern. I see two solutions: use the old name creation function for the functions that grpc-mqtt adds that are not in the .proto file, or just capitalize their names in the grpc-mqtt source so that both old and new versions of proto3-suite will produce the same compound name. I adopted the latter solution in ab569ab .

Edited to add that it seems the test failures I saw were different than those you saw. My change was focused on the test failures that I saw.