perf(c++): optimize primitive struct fields read performance (#2960)

## Why?

<!-- Describe the purpose of this PR. -->

## What does this PR do?

1. Type-based encoding detection: Added compile-time helpers to
correctly distinguish signed (varint) vs unsigned (fixed) integers:
- field_is_fixed_primitive<Index>() - bool, int8, uint8, int16, uint16,
uint32, uint64, float, double
- field_is_varint_primitive<Index>() - int32_t, int, int64_t, long long
(zigzag varint)
  2. Optimized fixed field reading:
- Compute field offsets at compile time with
compute_fixed_field_offset<T, I>()
- Read all fixed fields at absolute offsets without per-field
reader_index updates
    - Single reader_index update after all fixed fields
  3. Optimized varint field reading:
    - Track offset locally during batch reading
- Removed overly conservative max-varint-bytes pre-check (varints are
variable-length)
    - Single reader_index update after all varints
  4. Three-phase deserialization:
    - Phase 1: Batch read leading fixed-size primitives
    - Phase 2: Batch read consecutive varint primitives
    - Phase 3: Read remaining fields normally


## Related issues

#2958 
#2906 

## Does this PR introduce any user-facing change?

<!--
If any user-facing interface changes, please [open an
issue](https://github.com/apache/fory/issues/new/choose) describing the
need to do so and update the document if necessary.

Delete section if not applicable.
-->

- [ ] Does this PR introduce any public API change?
- [ ] Does this PR introduce any binary protocol compatibility change?

## Benchmark

| Datatype | Operation | Fory (ns) | Protobuf (ns) | Faster |
|----------|-----------|-----------|---------------|--------|
| Sample | Serialize | 103.9 | 59.2 | Protobuf (1.8x) |
| Sample | Deserialize | 329.3 | 478.1 | Fory (1.5x) |
| Struct | Serialize | 10.3 | 20.1 | Fory (1.9x) |
| Struct | Deserialize | 19.1 | 16.0 | Protobuf (1.2x) |

Author: chaokunyang