Field-aware member ordering in generated builders #328

Motivation

Methods in generated builders currently sort only by a global ordering bucket, then name, then signature. Grouping all direct setters at the end of the fluent section (done in #325 so MapStruct's last-wins accessor resolution binds the real setter) disconnected each field's methods: name(String) no longer sits next to name(Supplier), nameUpdate, add2Nicknames, etc.

Proposal: full member ordering

Order generated members by:

1. Constructors first
2. Access modifier — public before non-public
3. Scope — static members last
4. Field — methods belonging to the same bean field stay together (fields in a stable order, e.g. alphabetical as in the class-javadoc example chain)
5. Name — within a field, overloads of one name stay adjacent
6. Existing intra-name tie-breaks (parameter count, generic vs non-generic, signature)

With this layout each field's cluster reads naturally (name(String), name(Supplier), name(String,Object...), nameUpdate(...), add2Name(...)) while still guaranteeing the direct setter is the last overload of its name — which keeps #300's MapStruct binding working without the all-setters-at-the-end tradeoff.

Precedent

The class-javadoc example chain already groups exactly this way: fields alphabetical, each field's calls in a deterministic order (BuilderJavadocExampleTest asserts .pages(...), then .tags(...), then .title(...)). The member ordering would align the emitted class with what the javadoc already presents.

Implementation notes

• MethodDto today deliberately carries no field-origin metadata — the comparator would need a field/group key (populated in BuilderToGenerationTypeMapper from BuilderMethodDto.sourceFieldName), or a redesigned comparator in the rendering layer.
• Within a same-name overload group the direct setter must remain last for last-wins consumers (MapStruct, see MapStruct integration: ship a BuilderProvider SPI so generated builders are auto-detected #300).
• Regenerates example/generated-example-builder/** and the golden test expectations — large but mechanical diff.

Refs #300, #325