Cross compile tools for ARM (aarch64)

[echiugoog]

Cross compile supporting toolchain (where possible)
For perf, I bumped up the version to get a newer version of perf that supports Neoverse V2 cores

processwatch has ARM support, but we can't crosscompile right now as we need an ARM kernel to do the btf dump
e.g.

src/bpf/insn/build.sh
${BPFTOOL} btf dump file /sys/kernel/btf/vmlinux format c > ${DIR}/vmlinux.h

Did a basic perf test on a Neoverse V2 guest to validate perf:

$ sudo ./perf stat -- sleep 1

 Performance counter stats for 'sleep 1':

              0.53 msec task-clock                       #    0.001 CPUs utilized
                 2      context-switches                 #    3.784 K/sec
                 0      cpu-migrations                   #    0.000 /sec
                65      page-faults                      #  122.975 K/sec
           1547407      instructions                     #    1.10  insn per cycle
                                                  #    0.34  stalled cycles per insn
           1412410      cycles                           #    2.672 GHz
            504472      stalled-cycles-frontend          #   35.72% frontend cycles idle
            519564      stalled-cycles-backend           #   36.79% backend cycles idle
              8354      branch-misses

[harp-intel]

Looks awesome. One issue with where to get the aarch64 binaries when building with the container.

[harp-intel]

Too bad turbostat is x86 only. It's really useful for telemetry.

[harp-intel]

Like I said in the comment. This is very close. If I could make this one change myself I would do it and then merge, but I think you need to do it. Then, I can run my internal tests. I don't see any reason they would fail from these changes. Once passed, I can merge.
Thanks!

[harp-intel]

So close! This issue is also preventing my internal tests from running. Instead of, or in addition to, this change...

It would work best if the ARM build was in a separate artifact. So, the upload perfspect step (line 22) needs to be modified to only include perfspect.tgz. And, add an additional step to upload the ARM build.

[echiugoog]

attempted fix in c3f53cc

[harp-intel]

@echiugoog Thank you. This is awesome work. Internal tests on the x86 build have passed. I'm going to merge this after I publish the 3.8 release. Then I'll make a couple of tweaks to fix the test workflow in this repo.