Microoptimise SetContent

[Bios-Marcel]

• Remove reflect.DeepEqual call in SetContent

  There's no need for reflection as we know the types.
  Given tcell wants to stay compatible with older go versions, we now have
  a small function that handles this, instead of using something like
  slices.Equal.

• Add Simple benchmark to compare against the old versions
• Reuse currComb and lastComb slices to prevent allocations
• Optimise SetDirty calls in SetContent, don't re-lookup first cell

  This should be an improvent for ascii only rendering at least

Benchmark results:

D:\code\tcell> go test -bench=. -benchtime=10s .
goos: windows
goarch: amd64
pkg: github.com/gdamore/tcell/v2
cpu: AMD Ryzen 7 9700X 8-Core Processor
Benchmark_SetContentOld_ascii-16          108651            109888 ns/op
Benchmark_SetContent_ascii-16             116324            103552 ns/op
Benchmark_SetContentOld-16                 37220            323573 ns/op
Benchmark_SetContent-16                    55806            215194 ns/op
PASS
ok      github.com/gdamore/tcell/v2     55.717s