perf: allow multiple DATA frames per write

[NedAnd1]

Summary

This PR addresses the sender-side perf issue described in #902
by allowing multiple DATA frames to be coalesced into a single write_vectored() syscall,
rather than the current behavior of flushing each DATA frame individually.

Problem

The existing encoder holds at most one pending DATA frame at a time and blocks accepting new frames until it is fully flushed. With TCP_NODELAY enabled, each frame triggers its own write() syscall and TCP segment — inflating the syscall-to-payload ratio and cutting throughput roughly in half compared to raw TCP.

Solution

• Replace the single next Option<Next<B>> slot with a VecDeque<BufElement<B>> that can queue multiple DATA frames (up to 512 on vectored-IO transports).
• Implement Buf directly on the Encoder, providing chunks_vectored() so that all queued frame headers + payloads are written in one poll_write_vectored() call.
• Replace the per-connection in_flight_data_frame field with a per-stream in_flight_partial_send field of type Option<ControlFlow<()>>, allowing each stream to send up to one partial data frame at a time without blocking other streams.
• Introduce take_used_data_frames() iterator to reclaim all fully-written frames in a single pass, replacing the old take_last_data_frame() which could only return one.
• Add a custom poll_write_buf that returns ControlFlow to cleanly distinguish "nothing left to write" from "wrote some bytes, keep going".

Validation

• h2 cargo tests
• external scenario tests
• external perf tests

Perf results for a branch including the commit in this PR:

   OS: linux (kernel 6.6), Arch: arm64, TCP_NODELAY: True

   Data Frame Size │ master (throughput) │ batch-data-frames (throughput) │ Improvement
      4 kiB        │ 0.97 GB/s           │ 1.14 GB/s                      │ 1.18x
     16 kiB        │ 2.68 GB/s           │ 3.47 GB/s                      │ 1.29x
     64 kiB        │ 3.54 GB/s           │ 6.19 GB/s                      │ 1.75x
    256 kiB        │ 4.64 GB/s           │ 7.50 GB/s                      │ 1.62x
   1024 kiB        │ 4.28 GB/s           │ 7.81 GB/s                      │ 1.82x

Receiver-side PR: #904