Optimize Field creation when the SelectionSet doesn't contain fragments

[kyri-petrou]

During Field creation, we use a LinkedHashMap in order to deduplicate and combine fields depending on the field aliased name and the condition. I have no idea why this never occurred to me before, but this is only necessary when the SelectionSet contains fragments. In cases that it doesn't contain any fragments, we can simply just add just the Fields in a ListBuffer which is much cheaper.

Based on the newly added benchmarks, we see a 50-90% improvement in this post-validation phase.

PS: This phase was quite fast already, so I don't think we'll see much of this improvement reflected to the total execution time, but it's still nice not having to waste CPU cycles especially for services deployed in a k8s cluster

series/2.x:

[info] Benchmark                                         Mode  Cnt        Score       Error  Units
[info] ValidationBenchmark.postValidationDeep           thrpt   10  1919896.457 ± 16097.774  ops/s
[info] ValidationBenchmark.postValidationIntrospection  thrpt   10   227253.553 ±  1415.620  ops/s
[info] ValidationBenchmark.postValidationMultifield     thrpt   10  2223595.997 ± 12263.328  ops/s
[info] ValidationBenchmark.postValidationSimple         thrpt   10  5760613.925 ± 88803.131  ops/s

PR:

[info] Benchmark                                         Mode  Cnt        Score        Error  Units
[info] ValidationBenchmark.postValidationDeep           thrpt   10  3570988.986 ± 251199.323  ops/s
[info] ValidationBenchmark.postValidationIntrospection  thrpt   10   332137.620 ±  10887.654  ops/s
[info] ValidationBenchmark.postValidationMultifield     thrpt   10  4128116.557 ±   8987.987  ops/s
[info] ValidationBenchmark.postValidationSimple         thrpt   10  9185483.477 ±  80255.025  ops/s