The calibration modal was not starting properly, because
the component was not mounted when its show method
was called. Calling it from the `mounted` method solves
the problem and we don't get the error.
I'm not 100% sure this was introduced by refactoring the
taskSubmitter, though it's possible. However, I think the
fix is uncontroversial and it should probably have been there
all along.
I removed some defunct properties - this introduced a couple
of problems, which are resolved by:
* no longer checking for fastAutofocusUri
* removing the unused prop availablePlugins from the calibration modal
I've changed tasksubmitter to accept a "thing" and "action" rather than a single URL. This seems more in keeping with
propertyControl, and eliminates the need for loads of
computed properties.
I've also deleted (rather than just commenting out) chunks of the web app that are no longer in use. These may be
reinstated in the future - but we can get them from git
history. It's less confusing not to have vestigial
code in the repo.
This builds OK but is not, as yet, tested with hardware.
Previously, zip files were only downloaded if they already
existed. Now, they are created with an action, then
downloaded from the result.
Chrome now insists on downloading the files as `output.zip`, I can't see why this has changed, but perhaps it's related
to the security warning. This seems to be specific to zip files.
Stitching is now done automatically in a sub process by the
server. This is much cleaner - and all we need to do is poll for
file updates.
I've also styled the log display slightly better.
checkExistingTasks was looking up the task's `self` link by
checking for `rel=self`, which failed. I have corrected this
to `rel="self"` - `self` would have evaluated to `undefined`.
The log viewer was scrolling to the bottom when new logs
appeared, but not when the status updated. This meant that the status update may have been invisible. I now scroll
when the status changes too, which should fix the problem.
