Proper fix for error 'object has no attribute 'getSlave' in EtherCAT extension

traceback:
File "/home/developer/WorkData/PLC/beremiz/beremiz/IDEFrame.py", line 1433, in OnPouSelectedChanged
window.RefreshView()
File "/home/developer/WorkData/PLC/beremiz/beremiz/etherlab/ConfigEditor.py", line 837, in RefreshView
self.RefreshProcessVariables()
File "/home/developer/WorkData/PLC/beremiz/beremiz/etherlab/ConfigEditor.py", line 886, in RefreshProcessVariables
slaves = self.Controler.GetSlaves(**self.CurrentNodesFilter)
File "/home/developer/WorkData/PLC/beremiz/beremiz/etherlab/EthercatMaster.py", line 341, in GetSlaves
for slave in self.Config.getConfig().getSlave():
<type 'exceptions.AttributeError'>:_'lxml.etree._Element'_object_has_no_attribute_'getSlave'

Steps to reproduce problem:

- Add new EtherCAT master
- Add new EthercatNode to the master
- double click on


Revert commit "Dirty fix for error '_object_has_no_attribute_'getSlave' in EtherCAT extension"
[a3ac46366b86a0b237dac93be6b2281ac70b98a8].

The problem was that XML elements (proxy object) in some cases were created using custom XML
classes constructors and lxml.etree.Element() call and live python
patching. This causes that lxml backend doesn't know that custom python class
should be used for these XML elements.
Proxy object can be move/deleted and recreated by lxml
backend at any point in time or this can be done in python by copy/deepcopy operations.
If this happens, then newly created
proxy elements are using default class lxml.etree._Element. And all
custom functionality is lost.

All created XML elements should be always created through corresponding
parser and class lookup callback done by lxml backend.
It's described in more details in lxml documentation:
https://lxml.de/element_classes.html

How Do I Submit A Good Pull Request?
----------------------------------

It's highly recommended to write nice and clean python code. Beremiz
project tries to follows most of PEP-8 recommendations. They are
automatically checked on every push and merge by Bitbucket pipelines.

To avoid pushing "unclean" code, i's recommended to add one of the following
commands to pre commit Mercurial hook into .hg/hgrc configuration file.

Unfortunately script can't distinguish between real commit and shelve. If you
are using shelve (or maybe some other affected commands), it's recommended to
use pre-<command> and post-<command> hooks to create flags to skip checks on
some operations.


```
[hooks]
pre-shelve.linter = touch .hg/skiphook
post-shelve.linter = rm .hg/skiphook
pretxncommit.linter = ./tests/tools/check_source.sh --only-changes
```
or the same done using Docker container, so result will be the same as
on Bitbucket pipeline.

```
[hooks]
pre-shelve.linter = touch .hg/skiphook
post-shelve.linter = rm .hg/skiphook
pretxncommit.linter = hg status -m -n -a -n -I '**.py' --change $HG_NODE > files.lst && docker run --volume=$PWD:/beremiz --workdir="/beremiz" --volume=$PWD/../CanFestival-3:/CanFestival-3 --memory=1g --entrypoint=/beremiz/tests/tools/check_source.sh skvorl/beremiz-requirements --files-to-check files.lst
```