This includes dragging the libcxx implementation of normal_distribution
and uniform_int_distribution into our source. MSVC, clang and gcc don't
agree on how this should be implemented leading to problems with
comparing test results across platforms. Using a single implementation
gives the same results on each platform.

It uses a single QApplication object for the whole process rather than trying to
create an application for each test method

Fake dae responsive cancel

MonitorLiveData now cancels more quickly so the tests should not fail

This guarantees that background algorithms get
cancelled should StartLiveData raise an exception

Includes:
  - using stack allocation for tcp server
  - removal of unnecessary mutex
  - unused includes removed

Adding beam stop arm masking for all ISIS SANS instruments

If the source stops sending data then the cancellation
request has to wait until the receive loop completes
before it is checked. Add the check in more places
to improve the responsiveness.

Update OFFSPEC Parameter File

Add script to build dev site

Updated print statement to python 3 in SaveWorkspaces

Force higher compression xz format

ConjoinXRuns can receive input workspaces containing Dx values

Updated if statement in ExtractMask to work with MaskWorkspaces

- put each declaration of a variable in a single line
- using const auto  & to get x, y, e, dx
- directly set the size and xerrors in setPointStandardDeviations

Refs #22218

Fix  a date time string error in SaveVulcanGSS. 

SANS Reduced unit test file dependencies

Engineering Diffraction GUI - run GSASIIRefineFitPeaks in a separate thread

Added unit tests to userfile parser changes

Fix unreliable tests: Disable ClearCache doc test

(cherry picked from commit cbc85b74)

(cherry picked from commit 722870c1)

Re #20991

(cherry picked from commit 68806850)

Re #20991

(cherry picked from commit 673a5921)