Add btop-xe package for Intel xe GPUs

[stappmus]

Summary

Add a temporary btop-xe package that provides btop with Intel xe GPU support while upstream/Arch btop is still missing support for Xe-driver Intel GPUs.

This keeps the normal btop GPU build enabled (GPU_SUPPORT=true) and only swaps in a Xe-capable btop build on systems that explicitly install btop-xe.

Why

Intel Lunar Lake, Panther Lake, Battlemage, and future Intel GPUs use the xe kernel driver instead of i915. Current Arch extra/btop does not detect these GPUs, so Omarchy laptops with Intel Xe GPUs show no GPU usage/frequency in btop.

Upstream context:

• [Feature Request] Add Intel Xe driver support (Lunar Lake, etc.) aristocratos/btop#1407
• Add Intel Xe GPU driver support aristocratos/btop#1457

Package

• pkgname=btop-xe
• provides=('btop')
• conflicts=('btop')
• installs /usr/bin/btop
• restores cap_perfmon and cap_dac_read_search after install/upgrade
• pinned to the tested Xe-support commit 76530c80dd6184ccb72d7048c2589afdc4bdee52

Tested

Built locally with makepkg -f --clean --cleanbuild.

On Intel Panther Lake / Arc B390 using the xe driver:

/sys/class/drm/card0/device/uevent: DRIVER=xe
/sys/bus/event_source/devices/xe_0000_00_02.0 exists

After installing the same build locally:

btop version: 1.4.7+76530c8
Configured with: make STATIC= GPU_SUPPORT=true RSMI_STATIC=
/usr/bin/btop cap_dac_read_search,cap_perfmon=ep

btop -d reports:

Xe: Using fdinfo for GPU utilization (PCI: 0000:00:02.0)

GPU total usage, RC/MC engine usage, and frequency render in btop.

Notes

This package should be removed once upstream btop and Arch extra/btop include Intel xe support.

AI disclosure: this PR was prepared with AI assistance and manually tested on real Panther Lake hardware.

[stappmus]

Companion Omarchy wiring PR: basecamp/omarchy#6139

[stappmus]

Pushed an update to prefer Xe gtidle sysfs counters when the kernel exposes them, with fdinfo retained as the fallback for systems without gtidle.

On the Panther Lake/Xe laptop used for testing, the same 30s btop sample at update_ms = 2000 dropped from about 3.46% of one CPU core to about 1.13%, while still showing the embedded GPU row.