Use ReadOnlySpan<byte> for zero-allocation JSON parsing

[GmausDev]

Summary

Optimize JSON deserialization by using ReadOnlySpan<byte> and Utf8JsonReader to reduce string allocations during response parsing.

Current Behavior

var responseBody = await response.Content.ReadAsStringAsync(cancellationToken);
var result = JsonSerializer.Deserialize(responseBody, jsonTypeInfo);

This creates an intermediate string allocation for the entire response body.

Proposed Solution

Use stream-based or span-based deserialization:

// Option 1: Stream-based (preferred for large responses)
await using var stream = await response.Content.ReadAsStreamAsync(cancellationToken);
var result = await JsonSerializer.DeserializeAsync(stream, jsonTypeInfo, cancellationToken);

// Option 2: Span-based (for responses that fit in stack/pooled buffer)
var buffer = await response.Content.ReadAsByteArrayAsync(cancellationToken);
var result = JsonSerializer.Deserialize(buffer.AsSpan(), jsonTypeInfo);

Advanced: Pooled Buffers

For optimal performance, use ArrayPool<byte>:

var length = (int)response.Content.Headers.ContentLength!;
var buffer = ArrayPool<byte>.Shared.Rent(length);
try
{
    await response.Content.ReadAsByteArrayAsync(...);
    var result = JsonSerializer.Deserialize(buffer.AsSpan(0, length), jsonTypeInfo);
}
finally
{
    ArrayPool<byte>.Shared.Return(buffer);
}

Expected Benefits

• Reduced string allocations - No intermediate UTF-16 string
• Lower memory pressure - Direct UTF-8 parsing
• Faster parsing - Skip UTF-8 to UTF-16 conversion

Priority

🟢 P2 - Medium Impact

Technical Notes

• Stream-based is simpler and handles large responses well
• Span-based requires knowing content length upfront
• ArrayPool approach is most efficient but adds complexity
• Consider response size thresholds for different strategies