Poor rendering of stroke-dasharray with zero-length stroke segments (dots)

Migrated from: https://bugs.launchpad.net/inkscape/+bug/1568620

Steps to reproduce:

• open Inkscape
• open test file
• select paths one-by-one

svg	Inkscape

What happened?

Stroke segments with 0 length are only rendered (as dots) if they the first entry in the stroke-dasharray property.

What should have happened?

Dots should also be rendered if they are at the third, fifth, ... position of the array.

Inkscape Version and Operating System:

• Inkscape Version: 1.1-dev (12e2a09, 2020-02-19)
• Operating System: Windows 10
• Operating System version: 10.0.18362

Migration info

• Not reproduced with Inkscape 0.91 and trunk rev.14557 on the same computer.
• Reproduced with trunk rev. 14630. So it's probably due to a change between 14557 and 14630.
• Reproduced on Xubuntu 16.04, Inkscape 0.91 and trunk rev. 14987 (cairo 1.14.6).

The dashes passed to cairo and the calculated dash count are correct:

Message: Dash 0: 5,000000
Message: Dash 1: 5,000000
Message: Ndashes: 2
Message: Dash 0: 5,000000
Message: Dash 1: 1,250000
Message: Ndashes: 2
Message: Dash 0: 2,500000
Message: Dash 1: 1,250000
Message: Dash 2: 0,000000
Message: Dash 3: 1,250000
Message: Ndashes: 4
Message: Dash 0: 2,500000
Message: Dash 1: 1,250000
Message: Dash 2: 0,000000
Message: Dash 3: 1,250000
Message: Dash 4: 0,000000
Message: Dash 5: 1,250000
Message: Ndashes: 6
Message: Dash 0: 0,000000
Message: Dash 1: 5,000000
Message: Ndashes: 2

Adding 'g_message("Dash count: %i", cairo_get_dash_count(_cr));' after the cairo_set_dash call shows that Cairo only uses a dash array of size 2 in all cases (count from comment 5 where calculated from the array passed to cairo_set_dash).

To be tested with recent Debian based OSes (with Cairo 1.14 and Inkscape 0.91). If the error is in Cairo, then the stable version should be also affected.