refactor: use vim.loop.os_uname() for is_arm

[samirettali]

While configuring lazy loading in my config I noticed that copilot.lua took ~60ms on average to load, a lot of which is because of io.popen() so I changed the is_arm() function to use the .machine field of vim.loop.os_uname() which was already called.

I checked the neovim source code and i think it can still be simplified but I'm not entirely sure, this is the piece of code i found:

--- @param s 'x86_64'|'arm64'
--- @return boolean
function M.is_arch(s)
  if not (s == 'x86_64' or s == 'arm64') then
    error('unknown architecture: ' .. tostring(s))
  end
  return s == architecture
end

Load time with this change is ~13ms on average, I only tested it on an M1 Macbook.

[AntoineGS]

nice performance improvment :P