Search 6 groundwork - Restore support for bypassing bridges in the predicate/sort DSL

Description

Follow-up on HSEARCH-3221.

Essentially allow to do this in the predicate DSL:

.predicate().match().onRawField( "myField" ).orRawField( "myOtherField", "yetAnotherField" ).matching(<value>)

and this in the sort DSL:

.sort().byRawField( "fieldName" ).asc().onMissingValue().use( <value> )

... so that <value> is interpreted as the un-converted (raw) value of the field, and is not passed through the user-defined converter, but through a "pass-through" converter.

This should allow us to introduce a consistent syntax for the projection DSL, to project on a field without converting it:

.projection().rawField( "fieldName", String.class )

On the backend side, this means we will have to inform the backend, when fetching a builder, that:

Compatibility checks should ignore any custom converter
Methods such as org.hibernate.search.engine.search.predicate.spi.MatchPredicateBuilder#value, org.hibernate.search.engine.search.predicate.spi.RangePredicateBuilder#lowerLimit or org.hibernate.search.engine.search.dsl.sort.FieldSortMissingValueContext#use should ignore any custom converter and use a PassThroughToDocumentFieldValueConverter instead.

How to do that remains to be seen...