[havefunatcode]

FieldQueryMapEncoder and BeanQueryMapEncoder only used @Param on a @QueryMap POJO for the name, so expander was ignored:

@Param(value = "since", expander = DateToMillis.class)
private final Date since; // sent as Date.toString()

Both encoders now create the expander once per class, next to the existing metadata cache, and apply it in encode. It's instantiated with newInstance() like method params are, and a failure throws EncodeException. An Iterable is expanded per element (null elements stay null), anything else (arrays too) goes to the expander as a whole, same as a method @Param. If the expander returns null the entry is dropped like a null field.

Also fixed @Param(expander = Foo.class) with no value, which sent the param with an empty name. It falls back to the field/property name now.

@HeaderMap POJOs use the same encoder, so header fields get their expanders too. Header POJOs without an expander are unchanged. Happy to limit it to query maps if you'd prefer.

Added defaultEncoder_withParamExpander to both encoder tests (they fail on master) and mvn -pl core test passes. README and CHANGELOG updated.

Fixes #1312