nvx-rs slimming, phase 5.1: NVXC v2 replaces v1 #1089

Parent: #1076. Phase: 5, NVXC v2, the docs and the release. Needs: #1083, #1085, #1088. Unblocks: inside phase 5, groups of #1090.

Goal. nvx-init, the CLI and the fake OpenVMM speak NVXC v2 (#821), which streams an exec's input and output, and v1 and FEATURES are deleted. V2 lives in NVXC's private modules and has no compatibility promise.

Today (dev e42b1795) and after:

Groups, in order. Each group issue lists its PRs, with their goals, contents, Needs and estimates.

Notes.

• EXEC's payload is unchanged: the arguments, the working directory, the environment and the timeout.
• The guest stops reading a child's pipes while its send queue is full, so a slow host slows the workload instead of losing output.
• The deletion is leaving code. Its group issue opens with the note, and has no size limit.

Tests. Codec tests of Ready's protocol version and of each v2 frame; the host accepts the current protocol version and refuses another. The guest loop: a second EXEC gets BUSY; CANCEL and STOP work during an exec; stdin ends explicitly, each acknowledgement follows the write into the child's pipe, and at most 64 KiB is outstanding; the child's pipes aren't read while the send queue is full. The host session and the CLI: send_stdin, close_stdin, stdout and stderr streamed past the old 1 MiB cap, STOP during an exec, sandbox exec, sandbox run, and Ctrl-C sending CANCEL. The fake OpenVMM serves the same v2 frames to the CLI's fake-backed tests. Capture and restore, through the private restore driver and the live session that it returns, work over v2 before v1 goes. After the last group, no v1, FEATURES or GuestFeatures code or caller remains.

Done when: every group issue is closed; every NVXC path of NVX's host, guest and fake OpenVMM uses v2; and v1's frames, FEATURES and GuestFeatures are gone.