extreme_filter: performance gap between UInt8 and Gray{N0f8} for diamond shape SE

[johnnychen94]

After #102:

img = float64.(imresize(testimage("cameraman"), (512, 512)));
@btime erode($img, $(strel_diamond(img)));

img_nf08 = n0f8.(imresize(testimage("cameraman"), (256, 256)));
@btime erode($img_nf08, $(strel_diamond(img_nf08)));

type	size	time (ms)
UInt8	(256, 256)	0.008
UInt8	(512, 512)	0.028
UInt8	(1024, 1024)	0.108
Float64	(256, 256)	0.047
Float64	(512, 512)	0.251
Float64	(1024, 1024)	1.450

type	size	time (ms)
Gray{N0f8}	(256, 256)	0.014
Gray{N0f8}	(512, 512)	0.036
Gray{N0f8}	(1024, 1024)	0.118
Gray{Float64}	(256, 256)	0.049
Gray{Float64}	(512, 512)	0.247
Gray{Float64}	(1024, 1024)	1.487

This is mainly because rawview(channelview(img_n0f8)) is a MappedArray and should be fixed somewhere upstream (ImageCore, probably)

julia> versioninfo()
Julia Version 1.9.0-DEV.843
Commit 4873773d37e (2022-06-23 18:35 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin21.5.0)
  CPU: 16 × Intel(R) Core(TM) i9-9880H CPU @ 2.30GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-14.0.5 (ORCJIT, skylake)
  Threads: 8 on 8 virtual cores
Environment:
  JULIA_NUM_THREADS = 8