MVP golden path is missing 2 of its 8 commands: ebuild test and ebuild monitor

[srpatcha]

§7 names eBuild's MVP responsibility as:

eBuild — Setup, configure, build, test, flash, monitor.

§7.1 gives the golden path verbatim:

ebuild setup
ebuild new hello-eos
cd hello-eos
ebuild configure --board <reference-board>
ebuild build
ebuild test
ebuild flash
ebuild monitor

Success criterion: a new developer should reach a verified first firmware run
quickly and repeatably.

Two of the eight commands do not exist:

ebuild setup      EXISTS
ebuild new        EXISTS
ebuild configure  EXISTS
ebuild build      EXISTS
ebuild test       MISSING
ebuild flash      EXISTS
ebuild monitor    MISSING

Full command list today:

add analyze build clean configure firmware flash generate-board generate-boot
generate-project info install integration list-packages models new package
pipeline qemu repos sdk setup system

Twenty-three commands, and neither of the two the MVP definition names.

What this costs

The golden path stops at step 5 of 8. test is the step that makes the run
verified rather than merely completed, which is the stated success criterion —
without it a developer builds, flashes, and then has no way to find out whether
what they flashed works. monitor is how they would see it run at all.

ebuild qemu exists and is adjacent to monitor, but it is a different thing:
it launches an emulator, not a serial console onto whatever the developer just
flashed. Step 7 flashes real hardware; step 8 has to be able to watch real
hardware.

Scope

Two commands, following the CLI conventions already in ebuild/cli/commands.py:

• ebuild test — build the project's tests and run them, reporting the
  counts the underlying runner printed rather than inferring them from exit
  status. Must fail when tests fail; a test command that cannot go red is worse
  than none.
• ebuild monitor — open a serial console on the target's port, with the
  baud rate from the board configuration, --port and --baud overrides, and a
  clean exit on Ctrl-C. Must report an actionable error when no device is
  present rather than hanging.

Both need to work without hardware attached to be testable in CI: test runs on
the host, and monitor must fail cleanly and immediately with no port.