ArrowArrayAppendBytes has a narrowing cast when appending views with >INT32_MAX bytes

Found by Codex when reviewing https://github.com/apache/arrow-nanoarrow/pull/930

The offending narrowing cast is:

```c
if (private_data->storage_type == NANOARROW_TYPE_STRING_VIEW ||
private_data->storage_type == NANOARROW_TYPE_BINARY_VIEW) {
struct ArrowBuffer* data_buffer = ArrowArrayBuffer(array, 1);
union ArrowBinaryView bvt;
bvt.inlined.size = (int32_t)value.size_bytes;

if (value.size_bytes <= NANOARROW_BINARY_VIEW_INLINE_SIZE) {
memcpy(bvt.inlined.data, value.data.as_char, value.size_bytes);
memset(bvt.inlined.data + bvt.inlined.size, 0,
NANOARROW_BINARY_VIEW_INLINE_SIZE - bvt.inlined.size);
}
```

Start at the ArrowArrayAppendBytes entry point and inspect the string-view and binary-view handling shown in the issue. Trace how sizes are represented for values larger than INT32_MAX, then verify that appending such a view no longer performs a narrowing conversion while existing inline-view behavior remains correct.

Written by the indexing model from the issue text.