perf(hash): optimize hash function

[EdenEast]

This pr optimizes the hash function used to compute checksums for cached values. This is achieved by serializing a table into a minimum string representation and then performing the dj2b hashing algorithm on this string.

This pr optimizes the dj2b function by removing { string.byte } because it is only faster when the string is very long, which isn't the case for configurations

Benchmarks were on both main and perf branch.

local t = require("nightfox.config").options
local hash = require("nightfox.lib.hash")

local function bench(msg, f, iter)
  iter = iter or 1000
  local sum = 0
  for _ = 1, iter do
    local start = vim.loop.hrtime()
    f()
    sum = sum + (vim.loop.hrtime() - start)
  end
  print(msg, sum / iter / 1000000)
end

bench("perf", function()
  hash(t)
end, 10000)

Results

branch	time ms
main	0.4782018948
perf	0.0033633242

[EdenEast]

There is an issue with this solution as it is not reproducible over different lua sessions.

[nullchilly]

We can replace this PR with catppuccin/nvim#410