fix: honor IgnoreSetNullValue in specialized FieldReader subclasses, for issue #7618

[wenshao]

Summary

• The null-skip gate present in FieldReaderObject was missing from type-specialized FieldReader subclasses (FieldReaderString, boxed primitives, BigDecimal/BigInteger, Number, Date), so JSONReader.Feature.IgnoreSetNullValue was silently ignored for those fields — both when configured via @JSONType(deserializeFeatures = ...) and when passed at parse time.
• Adds ignoreSetNullValue() helpers on FieldReader so the gate is reused consistently, and applies it in accept(T, Object) and readFieldValue(JSONReader, T) for the affected specialized readers.
• ObjectReaderCreatorASM.genReadFieldValue:
  • extends the existing NullOnError fallback so it also falls back to fieldReader.readFieldValue when IgnoreSetNullValue is in fieldFeatures (annotation-driven case),
  • emits a runtime check on FEATURES for non-primitive fields on default-constructor beans, so the parse-time context feature also takes effect on the ASM hot path.
• The NoneDefaultConstructor ASM path uses temp locals (no OBJECT slot yet), so the runtime check is skipped there — that path needs a separate fix.

Fixes #7618.

Test plan

• New Issue7618 test covering all four combinations: String / boxed primitives × annotation-level / context-level feature.
• Existing related tests (Issue3789, Issue1745, Issue2164) still pass.
• All 2189 Issue* regression tests pass.
• All FieldReader* and ObjectReader*Test tests pass.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

wenshao seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}