Resolve "PipelineMemoryPrintEstimator stops at ImageList"

See merge request orfeotoolbox/otb!9

(cherry picked from commit 51bbe2d9)

0d1911c0 BUG: Track down upstream pipeline for each image in ImageList (fix #1516)
c47e4540 COMP: Fix compilation error (prototype mismatch)
3e227125 BUG: Moving PropageRequestedRegion() at the correct place in ImageList, and…
14c23e6b BUG: Only call UpdateOutputData() if object has Source

BUG: mantis 1504 - fix compilation error with itk 4.13 related to changes in GetDisplacementField signature

This reverts commit 8c843e0f.

The itk::TimeProbe constructor ends up calling
itkResourceProbe::GetSystemInformation, which reads things like the
hostname, the number of CPUs and so on. On Linux it parses /proc/cpuinfo
for that information.

OTB doesn't really need all those fields for a simple timer that's only
used in a couple of places. This commit adds a replacement for that.

As for the benchmarks, 100 000 instantiations of itk::TimeProbe take
about 9.5 s on my computer. I suspect it's somewhat slower on a server
with more CPUs. So this doesn't bring a huge performance improvement,
but it might help in cases where a large number of high-resolution
images are being read. The time for the same number of otb::Stopwatch
instantiations is 4 ms.

The context for this change is a project where we process time series of
medium to high resolution images over large areas. The TimeProbe in
otbGDALImageIO was sometimes showing in the profiles.

A relevant discussion can be found at
https://groups.google.com/d/msg/otb-developers/IvsFzSgRojo/0sMeEIZgCQAJ .

add Get and Set SignedSpacing in image and vector image otb class
modify spacing utilisation in GDALIO

- update CMake minimum required version
- set the C++ standard
- set otb::Image constructor to =delete to test