Transcoding optimizations

[zZHorizonZz]

Motivation:

This PR optimizes some transcoding utility classes and added more test to path matcher test class.

[vietj]

I think we could write a JMH benchmark for this

[zZHorizonZz]

I have been thinking about benchmarks for gRPC in general because there are TechEmpower tests which are testing Vert.x, but we don't really have a comparison of vertx-grpc with, for example, the grpc-ecosystem version.

That beign said, I don't think this is a major optimization because in general this only reduces some unnecessary calls, and in PercentEncoding it just puts the reserved chars to BitSet away from switch (which might make it a bit slower than using switch, but at the same time this way it is more readable, I think). Also if there was something for percent encoding already I would use that but I didn't really find anything in google lib it only provides escapers and not unescapers.

[vietj]

here I am thinking about microbenchmarking the convertion without gRPC transport itself

[zZHorizonZz]

Yeah, I can look into that 👍

[vietj]

we might be surprised by having extra bytes copy happening

[zZHorizonZz]

Well, I have added some basic benchmarks (I used Vertx core for the setup) and I have run the tests with old and the new. For MessageWeaver class, the new version is faster, but PercentEncoding class is seems slower. I will try to look into that.

[vietj]

where is this copied from ?

[zZHorizonZz]

It is in referenced in line underneath it, basically RFC 6570 and RFC 3986 are the same, in terms of reserved characters as RFC 6570 builds upon RFC 3986 I think?, but I find RFC3986 having little bit more explanation.

[vietj]

did you implement it ?

[zZHorizonZz]

What do you mean?