When computing the vector containing the detectors to ROI, in the
case of pixel and tube, erase elements after using remove, both to
remove a warning and because it is not supposed to work without this
change, even though it did.

It is now similar to the one from the Pick tab.

Add region/cut selection tool to the slice viewer

Colour fill plots fail on workspaces of monitors

Code can be the same whether the axis type is numeric/spectra
by using the correct algorithms

If an algorithm fails midway but we leave the workspace
around it looks like the operation completed causing
confusion for the user.

This commit adds checks for spectra labels being text and replaces them with arrays of length nhist

Add F(q) default parameter estimations

Add Setting for Default Font on New Plots

Adds the possibility to select and group individual pixel and tube.
HOWEVER due to these selections being dependent of the detector, contrary
to every other shape used for masking etc, the selection cannot be
printed on screen, as it would require to create a shape that follows
the detector if it is moved and is rendered using the correct
perspective.
Moreover, contrary to what happens with masking and ROIing, there are
two options for grouping, which means the operation cannot be applied
immediatly on click in this case.
Since this means that a) the user should input after doing their selection
and b) this selection cannot be shown to the user on the instrument
view, the compromise solution was to only take into account and store
the last tube or pixel selected for grouping. In other words, the user
can only sum or extract the last pixel / tube they selected.

RE #28597

Muon Analysis bug fixes for simultaneous fits

Abins data class refactoring

The same key combinations, minus roi selection, now export
the visible cuts when the single-pixel line plots are
visible.

It is highly likely that this is the behaviour that is
intended rather than simply throwing an error. If makes
extracting a single bin out far simpler than a user
figuring out the correct limits themselves.

RE #28597

The toolbar was incorrectly stretched over 2
columns but only macOS show this as a layout
error.
Also, macOS buttons were sized far too small
and the text was cut off. Calculate the width
of the buttons based on their height so they
appear square

Fix macOS layout properly

Some values cannot be displayed so show a blank

Some units are not convertible so we just show a dash to
show it's not applicable.

We only want to borrow a reference to the PyObject for ExtractSharedPtr
to work. The handle objects are able to do this but providing access to
the raw PyObject in ExtractSharedPtr avoids the overhead of constructing
the handle when not required.

Adds more tests cases for common MatrixWorkspace setups

is grouped then the recursive lookup fails even if the instrument
has the required parameter

Adds a status bar to inform the user what keys to press and
what has happened when the keys have been pressed.
It remains to be seen if separate push buttons are required
at all here.

As we now view the original workspace if possible rebinning
happens automatically with a new plot

The QuadMesh cannot currently provide the information
to compute its position in the grid. Instead just
display the x,y position as matplotlib does.

When lineplots are enabled the image info widget
and lineplots need the same information and this
avoids computing it twice.

Use separate objects to manage tracking and plotting
to enable better testing and readability.