* Instead of Person.reference, Person.user_id is now used to log in (fe27a9c3).
* Fields text were renamed to follow naming conventions (329c196d).

/reviewed-on !770

/reviewed-on !763

This snapshot were made on Chrome 70 (ChromeOS) with a 1001x932 window
(innerHeight x innerWidth)

To Compare a "screenshot" of a DOM element with a reference snapshot.

This check supports the following parameters:
 *   locator - an element locator
 *   misMatchTolerance - the percentage of mismatch allowed. If this is 0, the
      images must be exactly same. If more than 0, image will also be resized.

https://github.com/niklasvh/html2canvas Copyright (c) 2012 Niklas von Hertzen

from https://github.com/HuddleEng/Resemble.js/blob/v2.10.2/resemble.js © 2013 Huddle

to get stable sort and prevent "random" test failures.

"Restore" sounds a bit strange to me in this context.

So that we can wait for this in the tests

[erp5_officejs] Use upload_dict instead of file_extension to check for correct file-type during notebook upload

Also, exlicitly specify the txt extension during download

/reviewed-on !760

/reviewed-on !762

Also, update tests for notebook to check for Upload page

/reviewed-on !764

Fixes rendering when using Chameleon:
  ValueError: only strings can be translated, not: <ERP5Type.Message.Message for ...

/reviewed-on !747

/reviewed-on !765

adjust font url

This is incubator-echarts-4.1.0-release/dist/echarts.min.js from
https://github.com/apache/incubator-echarts/releases/tag/4.1.0-release

https://www.erp5.com/documentation/developer/guideline/coding/erp5-Guideline.Coding.Best.Practice#use-methods-to-access-properties-in-erp5

Otherwise filling cache fail and it's never used.

/reviewed-on !759

This reapplies parts of 9a84ef83 , but a
different server side approach:
 * use portal_workflow API to count worklists
 * introspect workflow to build queries
 * use translated state in the query, trying to be a bit more user
   friendly. We could do for portal_type as well.

This workarounds #20180907-24A654C a problem with queries generated by
gadget worklist does not seem to support complex queries with multiple
values ( local_roles: ['Assignor', 'Assignee'] ). This code generates
queries:

```
( ( local_roles:  "Assignee" OR local_roles:  "Assignor" ) AND portal_type:  "Support Request" AND translated_simulation_state_title:  "Opened" )
```

where the previous version generated queries like:

```
( portal_type:  "Support Request" AND simulation_state:  "validated" AND local_roles: (  "Assignee" OR  "Assignor" ) )
```

which did not match anything. Maybe because local_roles is handled
differently in catalog