Bytecode rewriting for reduced event dispatching overhead

[SirYwell]

Draft because there are many moving parts here. Comments and feedback are welcome though.

See #14153, especially #14153 (comment).

The bytecode rewriting/gradle integration is probably problematic regarding caching.

The current micro benchmark shows the following numbers on my machine:

Benchmark                                  (hasListener)  Mode  Cnt  Score   Error  Units
EventBenchmarks.callNormal                          true  avgt    6  6.507 ± 0.345  ns/op
EventBenchmarks.callNormal                         false  avgt    6  0.124 ± 0.010  ns/op
EventBenchmarks.hasListenersIndy                    true  avgt    6  0.496 ± 0.030  ns/op
EventBenchmarks.hasListenersIndy                   false  avgt    6  0.495 ± 0.038  ns/op
EventBenchmarks.registeredListenersLength           true  avgt    6  0.626 ± 0.027  ns/op
EventBenchmarks.registeredListenersLength          false  avgt    6  0.613 ± 0.024  ns/op