Use ratio when calculating ImageWMS width and height

[ahocevar]

Fixes #4403.

[tschaub]

This looks like it corrects the single image WMS example (on a non-retina device at least). It would be great to get in a regression test for this, but I think a patch release is more urgent.

Thanks for the fix. Please merge.

[ahocevar]

I was planning to add tests for this anyway, but decided to look into #4405, the other serious regression, first.

[ahocevar]

BTW, fixed for Retina displays too.

[tschaub]

Thanks @ahocevar.

The downside of using a branch as the base for patches is that the continuous-integration/travis-ci/pr job doesn't run. I just pushed a change to our Travis config to make the continuous-integration/travis/push job run for branches named like /^v.*\.x$/. Not as convenient since we don't get the commit status updated on a PR. But we can check the build status periodically (e.g. https://travis-ci.org/openlayers/ol3/builds/90780614).

The alternative is to commit patches to master and then cherry-pick. But that brings potential for conflicts.

Update - as an alternative to triggering the Travis push job for branches named like v.X.Y.x, I've just opened a pull request for this branch (see #4410). There we can watch the commit status.