feat(java): add deserialize overloads that take a TypeRef (#3946)

[Swatantra-66]

Why?

ForyJson has fromJson methods that accept a TypeRef for cases when the root type is generic (e.g. List<T>, Map<K, V>). However, core Fory previously only provided deserialize overloads accepting Class<T>.

Because Java type erasure strips generic type arguments at runtime from Class, deserializing into generic root collections or maps lost element and key/value type information. Users were unable to deserialize directly into generic root types without attempting low-level, internal TypeResolver manipulations.

What does this PR do?

Adds TypeRef<T> deserialize overloads across BaseFory, Fory, ThreadLocalFory, ThreadPoolFory, and BlockedStreamUtils:

1. API Declarations (BaseFory):

   • <T> T deserialize(byte[] bytes, TypeRef<T> typeRef)
   • <T> T deserialize(MemoryBuffer buffer, TypeRef<T> typeRef)
   • <T> T deserialize(ForyInputStream inputStream, TypeRef<T> typeRef)
   • <T> T deserialize(ForyReadableChannel channel, TypeRef<T> typeRef)

2. Core Implementation (Fory):

   • Implemented null-checked overloads (Preconditions.checkNotNull(typeRef, "typeRef must not be null")).
   • Unified deserializeByType into a clean helper accepting GenericType and raw Class<?>, eliminating duplicate code. Pushes typeResolver.buildGenericType(typeRef) to readContext.getGenerics(), preserving root generic parameter metadata (e.g. element type for CollectionLikeSerializer, key/value types for MapLikeSerializer).

3. Thread-Safe Runtimes:

   • ThreadLocalFory: delegates all 4 overloads to currentFory().deserialize(..., typeRef).
   • ThreadPoolFory: delegates all 4 overloads with thread pool acquire() / release().

4. Streaming Helper (BlockedStreamUtils):

   • Added deserialize(Fory, InputStream, TypeRef<T>) and deserialize(Fory, ReadableByteChannel, TypeRef<T>).

5. Comprehensive Tests:

   • ForyTest.java: Added testDeserializeWithTypeRef testing root List<String>, nested Map<String, List<Integer>>, concrete objects via TypeRef.of(...), byte arrays, memory buffers, input streams, channels, and null checks.
   • ThreadSafeForyTest.java: Added testSerializeDeserializeWithTypeRef validating both ThreadLocalFory and ThreadPoolFory.
   • StreamTest.java: Added TypeRef tests for testStream and testReadableChannel.
   • BlockedStreamUtilsTest.java: Added stream and channel deserialization tests with TypeRef.

Related issues

Closes #3946

AI Contribution Checklist

• Substantial AI assistance was used in this PR: no

Does this PR introduce any user-facing change?

• Does this PR introduce any public API change? Adds deserialize overloads taking TypeRef<T>.
• Does this PR introduce any binary protocol compatibility change? No.

Benchmark

No performance impact on existing Class<T> paths. The unified deserializeByType helper avoids duplicate code and overhead.

[ayush00git]

i see a lot of issues in this PR, java/fory-core/src/main/java/org/apache/fory/Fory.java. The root reader now declares element types that the root writer never declared. your tests even not reach the path you changed.

[Swatantra-66]

Thanks a lot for the review and catching this, @ayush00git!

1. Root element types: Because fory.serialize(obj) has no generic context at the root level, the writer emits dynamic element headers/type tags. By pushing buildGenericType(typeRef) into readContext.getGenerics(), the reader was incorrectly expecting monomorphic elements without type headers on the wire. I have updated all deserialize(..., TypeRef<T>) overloads to delegate cleanly to (Class<T>) typeRef.getRawType(), avoiding wire protocol desynchronization while maintaining compile-time type safety for callers.
2. Test coverage: Arrays.asList() delegates to ArraysAsListSerializer (Object[]), bypassing CollectionLikeSerializer.readElements(). I've updated the tests to use concrete ArrayList, LinkedList, HashSet, HashMap, and nested generic structures, verifying that the actual collection serializers are exercised.

Pushed the updated commit!