fix(frecency)!: Use byte strings for DB IDs

[mrjones2014]

Resolves: #472

How to Test

1. Note that existing frecency data will unfortunately be invalid because the ID algorithm changed.
2. Try an item with a '' and/or " in the description e.g.

return {
    'kylechui/nvim-surround',
    version = "*", -- Use for stability; omit to use `main` branch for the latest features
    keys = {
        { "<leader>'", '<Cmd>normal cs"\'<CR>', desc = 'Change surrounding " to \'' },
        { '<leader>"', '<Cmd>normal cs\'"<CR>', desc = 'Change surrounding \' to "' },
    },
    config = function ()
        require('nvim-surround').setup({ })
    end
}

with lazy_nvim extension and the frecency sorting enabled.

Testing for Regressions

I have tested the following:

• Triggering keymaps from legendary.nvim in all modes (normal, insert, visual)
• Creating keymaps via legendary.nvim, then triggering via the keymap in all modes (normal, insert, visual)
• Triggering commands from legendary.nvim in all modes (normal, insert, visual)
• Creating commands via legendary.nvim, then running the command manually from the command line
• augroup/autocmds created through legendary.nvim work correctly