Add tinyglobby to product benchmark instead of fdir

[mrmlnc]

First, the fdir package is not a glob solution. Yes, it can be used as a glob, but it simply uses picomatch to find matching paths and does not apply any heuristics. This is clearly visible in our benchmarks. Rather, it is a competitor to @nodelib/fs.walk, not fast-glob.

Using ~/work/fast-glob/fast-glob/herebyfile.mjs to run bench:product:async
Starting bench:product:async:flatten
Label              import.time (time)  time (time)       memory (memory)    entries (value)  process.time (time)
-----------------  ------------------  ----------------  -----------------  ---------------  -------------------
async fast-glob *  10.146ms ±2.178%    4.414ms ±1.719%   5.389 MiB ±0.000%  6 ±0.000%        43.573ms ±1.645%   
async node-glob *  14.556ms ±6.243%    6.483ms ±3.063%   5.137 MiB ±0.293%  6 ±0.000%        50.223ms ±2.531%   
async fdir *       9.138ms ±8.759%     43.941ms ±4.199%  6.380 MiB ±7.375%  6 ±0.000%        85.082ms ±3.145%   

Finished bench:product:async:flatten in 35.8s
Starting bench:product:async:deep
Label               import.time (time)  time (time)        memory (memory)     entries (value)  process.time (time)
------------------  ------------------  -----------------  ------------------  ---------------  -------------------
async fast-glob **  10.139ms ±1.517%    67.479ms ±4.974%   10.663 MiB ±0.800%  17321 ±0.000%    109.259ms ±4.383%  
async node-glob **  14.292ms ±5.569%    159.467ms ±3.012%  25.656 MiB ±2.180%  17321 ±0.000%    209.310ms ±2.666%  
async fdir **       8.980ms ±8.204%     51.330ms ±2.906%   8.832 MiB ±1.093%   17321 ±0.000%    92.592ms ±1.898%   

Finished bench:product:async:deep in 1m 22.2s
Starting bench:product:async:partial_flatten
Label                                            import.time (time)  time (time)       memory (memory)    entries (value)  process.time (time)
-----------------------------------------------  ------------------  ----------------  -----------------  ---------------  -------------------
async fast-glob {fixtures,out}/{first,second}/*  10.243ms ±1.576%    8.927ms ±1.916%   5.044 MiB ±0.126%  2 ±0.000%        48.568ms ±1.513%   
async node-glob {fixtures,out}/{first,second}/*  14.351ms ±5.512%    8.226ms ±2.810%   5.264 MiB ±0.374%  2 ±0.000%        51.684ms ±2.104%   
async fdir {fixtures,out}/{first,second}/*       9.071ms ±8.824%     43.779ms ±3.488%  6.503 MiB ±6.242%  2 ±0.000%        86.557ms ±3.088%   

Finished bench:product:async:partial_flatten in 37.4s
Starting bench:product:async:partial_deep
Label                              import.time (time)  time (time)       memory (memory)    entries (value)  process.time (time)
---------------------------------  ------------------  ----------------  -----------------  ---------------  -------------------
async fast-glob {fixtures,out}/**  10.203ms ±4.545%    9.877ms ±5.598%   5.239 MiB ±0.085%  178 ±0.000%      48.916ms ±3.786%   
async node-glob {fixtures,out}/**  14.350ms ±5.594%    10.967ms ±2.434%  5.546 MiB ±0.295%  178 ±0.000%      54.478ms ±2.044%   
async fdir {fixtures,out}/**       9.029ms ±8.476%     43.141ms ±2.229%  6.492 MiB ±6.712%  178 ±0.000%      85.306ms ±1.786%   

Finished bench:product:async:partial_deep in 37.7s
Starting bench:product:async
Finished bench:product:async in 1ms
Completed bench:product:async in 3m 13.3s

Secondly, another glob package appeared in the ring, which is based on the fdir package. So, great! I'll stay here and observe.

I'm simply replacing fdir with tinyglobby.