Perf: use io copy with buffer

[Mmx233]

Using io.CopyBuffer can reduce the resource consumption caused by multiple memory allocations and appropriately increase the buffer size to optimize the performance of Copy.

This may mitigate the issue described in #6325

[Mmx233]

Test image available at mmx233/alist:v3.33.0-rc2, I will test it at tomorrow morning for my test environment is down now.

[Mmx233]

My test environment is keeping unavailable. I tested it on my local computer. Using a 20GB single file, Windows amd64 SSD, with signature enabled, HTTP service, Chrome single-threaded download, and Free Download Manager 15-threaded download.

Only the occupation of alist is being measured. Iowait is an important performance metric but cannot be calculated in the current environment. I only ran it once and manually retrieved the data from the task manager, and there may be many other unstable factors in between. So the resulting outcome should be considered for reference only.

original io copy

Threads	CPU Usage	Memorary Usage	Speed
1	10%	19.2Mb	400Mb/s
15	12%	20.9Mb	500Mb/s

io copy buffer

Threads	CPU Usage	Memorary Usage	Speed
1	4.5%	16.5Mb	400Mb/s
15	8%	19Mb	485Mb/s