Add developer VM script for faster testing iteration

[Softer]

PR Description

Sharing a small helper I've been using locally to iterate on archinstall changes inside a VM. It saved me enough time that I figured other contributors might want it too.

The pain

When I change something in archinstall and want to test it end-to-end (run the installer in a VM, walk the menus, let it actually partition and install), the current options all share the same shape.
build_iso.sh rebuilds the full ISO on every source change, carrying gcc/pip/uv/wheel build each time. The README "Testing" section and the wiki "Building and Testing" describe the same loop manually: boot
vanilla Arch ISO, pacman -S git python-pip gcc pkgconf, clone repo, run python -m archinstall. Both work, but both rebuild the world on every change.

What I wanted was: edit a .py on the host, relaunch the VM, run archinstall, see the change. No wheel build. No ISO rebuild. Just a hot-reload dev loop.

What this PR adds

dev_vm.sh at the repo root, sibling of build_iso.sh. Two independent pieces:

1. A minimal dev ISO built with archiso that carries only archinstall's runtime deps - no gcc, no pip, no wheel build. The dep list is derived at build time from pyproject.toml's [project.dependencies]
   (PEP 503 normalized, mapped to the Arch python-<name> convention, which holds for every current dep), so adding or removing a runtime dep upstream does not require touching the script. The ISO is tiny and
   only needs rebuilding when the runtime dep list actually changes.
2. A QEMU launcher that shares the project source read-only to the guest over 9p. Host edits appear live inside the VM. On guest login, the 9p share auto-mounts at /root/archinstall-dev, the shell cd-s
   into it, and archinstall is aliased to python -m archinstall from the shared source. A 9p mount failure prints a clear error, so a broken virtfs setup does not silently turn into "No module named
   archinstall".

So the inner loop becomes: edit code on host, ./dev_vm.sh k, type archinstall inside the guest. No rebuild, no reinstall.

Modes

• ./dev_vm.sh - build ISO if missing, fresh disk, boot
• ./dev_vm.sh rebuild | r - force ISO rebuild, fresh disk, boot
• ./dev_vm.sh keep | k - reuse disk, boot ISO
• ./dev_vm.sh boot | b - boot from installed disk (no ISO)
• ./dev_vm.sh clean | c - remove disk, NVRAM, ISO
• ./dev_vm.sh -h - help

Host artifacts (all git-ignored)

• .dev-iso/ - generated dev ISO
• .dev-disk.qcow2 - VM disk
• .dev-ovmf-vars.fd - persistent UEFI NVRAM
• .dev-configs/ - optional, user-created, shared rw to guest at /root/cfg (handy for keeping test user_configuration.json / user_credentials.json between runs)

Host requirements

Arch-based host (Arch, Manjaro, EndeavourOS, ...) because mkarchiso is Arch-only. The script reads /etc/os-release and fails fast with a clear message on non-Arch hosts, before ever touching pacman.
Runtime deps: qemu-base, edk2-ovmf, sudo, python (for parsing pyproject.toml). OVMF paths are probed across the common layouts (Arch edk2-ovmf, Debian/Ubuntu ovmf, Fedora edk2-ovmf) so the same
script works across Arch-based distros with different OVMF packaging. 9p mapped-xattr mode needs xattr support on the filesystem holding the repo (ext4/btrfs work out of the box; ZFS needs xattr=sa).

Changes in this PR

• dev_vm.sh - new file
• .gitignore - three lines: !dev_vm.sh as an exception under the existing **/**.sh blanket, plus /.dev-iso/ and /.dev-configs/ for the artifact folders

Open to feedback on placement (repo root vs a scripts/ folder), naming, and scope. I kept artifact names dot-prefixed to hide them from ls and group them under one visual namespace.

Note: the CI ISO artifact attached to this PR is the regular build_iso.sh output, unaffected by dev_vm.sh (which is a separate dev-only tool and does not touch the CI build path).