Changes v2 IDs and IPs to validated Strings

[codefromthecrypt]

This changes v2 IDs and IPs to validated Strings, which matches the open api definition http://zipkin.io/zipkin-api/#/

This also polishes codec and backfills the Annotation and Endpoint types missing from the WIP v2 model.

Benchmark results are about 2x efficient as prior:

Benchmark                                          Mode  Cnt   Score   Error  Units
CodecBenchmarks.readClientSpan_json_span2          avgt   15   4.220 ± 0.103  us/op
CodecBenchmarks.readClientSpan_json_zipkin         avgt   15   8.371 ± 0.404  us/op
CodecBenchmarks.readClientSpan_thrift_libthrift    avgt   15   2.020 ± 0.064  us/op
CodecBenchmarks.readClientSpan_thrift_zipkin       avgt   15   2.055 ± 0.051  us/op
CodecBenchmarks.readTenClientSpans_json_span2      avgt   15  29.917 ± 0.969  us/op
CodecBenchmarks.readTenClientSpans_json_zipkin     avgt   15  67.485 ± 2.951  us/op
CodecBenchmarks.readTenClientSpans_thrift_zipkin   avgt   15  21.180 ± 0.816  us/op
CodecBenchmarks.writeClientSpan_json_span2         avgt   15   1.000 ± 0.041  us/op
CodecBenchmarks.writeClientSpan_json_zipkin        avgt   15   1.916 ± 0.048  us/op
CodecBenchmarks.writeClientSpan_thrift_libthrift   avgt   15   2.860 ± 0.092  us/op
CodecBenchmarks.writeClientSpan_thrift_zipkin      avgt   15   0.660 ± 0.029  us/op
CodecBenchmarks.writeTenClientSpans_json_span2     avgt   15  10.249 ± 0.260  us/op
CodecBenchmarks.writeTenClientSpans_json_zipkin    avgt   15  19.509 ± 1.474  us/op
CodecBenchmarks.writeTenClientSpans_thrift_zipkin  avgt   15   6.326 ± 0.081  us/op

See #1499 (comment)