Should we do imshow(self, *kwargs) and allow users to pass extra keyword arguments to grdimage?

I'm a little on the fence about adding a colorbar. Sometimes the default colorbar looks very bad (e.g. too many black lines). Do you think we should:

1. Add a setting to imshow like show_colorbar=True to enable/disable the colorbar if needed?
2. Allow extra keyword arguments to colorbar somehow, so that users can tweak it if needed?

Or would this just not show a colorbar by default? What do others think?

I'm a little on the fence about adding a colorbar. Sometimes the default colorbar looks very bad (e.g. too many black lines).

That's because the original CPT file has too many z-slices. fig.colorbar(frame=True) should always be good.

Debating on whether we should add return fig, so that users could potentially add extra stuff to the figure like so:

fig = grid.gmt.imshow()
fig.colorbar()
fig.show()

or would that get a bit confusing? I'm thinking a bit about how matplotlib does it where users can add extra elements to the figure, but unsure if we want to mimic that in PyGMT.

See my comment in #1412 (comment).

Initially, I propose to add the imshow() function as a quick DataArray viewer, so I expect to view the grid using a simple single command.

Similar to the ax parameter in xarray.DataArray.plot.imshow, I think the imshow function can at least accept a fig parameter.

If fig=None (the default behavior), grid.gmt.imshow() should plot the grid and show the image.

If a Figure instance is given, then we know that users want to plot the grid in an existing figure, then we should not show the image.

Then the imshow() will be like:

    def imshow(self, fig=None):
        if fig is None:
            fig = Figure()

        fig.grdimage(self._obj, frame=True)
        fig.colorbar()

        if fig is None:
            fig.show()
        else:
            return fig