feat: support concurrent chunk uploads

[TorstenDittmann]

This PR updates the SDK to support concurrent chunk uploads.

[greptile-apps]

Greptile Summary

This PR refactors chunked file uploads to run up to 8 chunks concurrently using Interlocked counters and a resultLock for safe result aggregation. It also fixes a pre-existing threading bug where per-request headers were being written to the shared _http.DefaultRequestHeaders instead of the per-request request.Headers.

• Concurrent workers: Chunks are dispatched via Interlocked.Increment over a pre-built list; the first chunk is uploaded sequentially to establish the upload ID before parallel workers start, avoiding any race on uploadId.
• Header isolation fix: Moving header mutations from _http.DefaultRequestHeaders to request.Headers is a correct and important fix that prevents cross-request header pollution in concurrent call scenarios.
• File handle leak resolved: The previous info.OpenRead() assignment that left a dangling FileStream in input.Data has been removed; ReadChunkAsync now opens a fresh, properly-disposed stream per chunk for the \"path\" source type.

Confidence Score: 5/5

Safe to merge; the concurrency logic is well-structured and the header-isolation fix removes a latent cross-request mutation bug.

The first chunk is uploaded sequentially before workers start, so uploadId is written exactly once before any parallel reads. Worker coordination via Interlocked and resultLock is correct. The only concern is a synchronous lock for stream reads that blocks thread pool threads, but this is a design-quality issue rather than a correctness defect.

Appwrite/Client.cs — the stream-read locking pattern in ReadChunkAsync is worth revisiting for thread-pool-constrained hosts.

Important Files Changed

Filename	Overview
Appwrite/Client.cs	Adds concurrent chunk uploads with up to 8 parallel workers using Interlocked/lock primitives; also fixes a critical threading bug where per-request headers were being mutated on the shared _http.DefaultRequestHeaders. Stream-backed inputs serialize reads via a synchronous lock inside the async method, which can block thread pool threads.

_{Reviews (2): Last reviewed commit: "feat: support concurrent chunk uploads" | Re-trigger Greptile}