Construct the significant terms aggregation builder's parser statically

[nik9000]

Right now we build a new ObjectParser every time we want get a significant terms aggregation. We do this because we had to pass in ParseFieldRegistry<SignificanceHeuristicParser>. I think we can avoid this by moving ParseFieldRegistry<SignificanceHeuristicParser> to NamedXContentRegistry style parsing.

[nik9000]

Marked a low hanging fruit because it should be fairly simple to do but big enough that it certainly deserves its own PR.

[cbuescher]

Same is true for SignificantTextAggregationBuilder which follows the same pattern. I took a quick first look at it, changing the way the heuristics are registered is only part of this. I quickly got stuck when trying to convert the part around

elasticsearch/core/src/main/java/org/elasticsearch/search/aggregations/bucket/significant/SignificantTermsAggregationBuilder.java

Line 98 in f576c98

new ParseField(name));

because currently each potential heuristic is declared on the ObjectParser using its name. I think we might need something like what I started in #24020 to be able to do this with ObjectParser, most other places we parse named things like this we still do it in the "classic" manual parsing style. Just adding this note here since this issue might be a little more involved than estimated.

[colings86]

@elastic/es-search-aggs