Undo accidental revert

As noted in a comment in this BT's Base_redirect implementation, original code
does raise when abort_transaction is true.
Not raising in this implementation means that this script will return to caller,
while it never does on original code.
Also, to add insult to injury, this utterly bogus implementation interferes with
transaction boundaries. So suddenly, a single publication spans over 2
transactions, which can lead to:
- ZODB Connection sharing, breaking transaction isolation
- the second transaction implicitly created by this abort (actually, by the next
transactional connector registration to transaction) may be committed,
in which case anything done after Base_redirect returns will be
persistently committed, against caller's explicitly specified intent, and against
all developer expectations.

NEVER TOUCH TRANSACTION ! Only CMFActivity and unittests are allowed
this level of access (and CMFActivity should be modified out of this exceptional
state).

Linshare Use the same configuration than ERP5

Thanks to Vincent Bechu for his work on this topic.

This reverts commit fba5767e.

Sorry, I went too quickly, there was difference between self and context

This reverts commit 7aa5fb76

> this change makes title empty when reply
nexedi/erp5!823 (comment 73439)

/reviewed-on nexedi/erp5!826

1. a inventory report module is added which user can add inventory report


2. in inventory report,  user should put information such as Owner, Warehouse... then call calculate product stock action


3. one line per product is then added automatically with current total quantity and total asset price


![Screenshot_2018-12-12_at_17.42.24](/uploads/98f8a4af66faadfdfd1d99b4400ff687/Screenshot_2018-12-12_at_17.42.24.png)


/reviewed-on nexedi/erp5!806

if i have a url like:
https://erp5xxx/form_idxxx?relative_item_relative_url=item_module/7817

with the old way, form id will be 7817

/reviewed-on nexedi/erp5!824

According to pep-0263, python scripts can define encoding magic on
first or second line.
If we use for example such a script:

```python
  # coding: utf-8
  return "héhé"
```

then using .read() was wrong, because it return the script with headers,
ie:

```python
  ## Script (Python) "test_coding"
  ##bind container=container
  ##bind context=context
  ##bind namespace=
  ##bind script=script
  ##bind subpath=traverse_subpath
  ##parameters=
  ##title=
  ##
  # coding: utf-8
  return "héhé"
```

so the coding magic is no longer in first line and pylint complains
with error like:

Cannot decode using encoding "ascii", unexpected byte at position 11 (invalid-encoded-data)]

/reviewed-on !825

only calculate exchange rate if necessary
use transaction cache
make getExchangeRate as local function

remove useless test_21
update test_20 with various packing list

multiple owner test case
multiple cancel packing list test
circular movement test

This method is used to import ZODB Components when fallback on filesystem is needed
and used when loading Document and Extension Components.

1,000,000 executions:
  * Before:
    timeit("erp5.component.extension.find_load_module(module_name)",
           setup="import erp5.component.extension; module_name='Hoge'")
    59.073
    59.324
  * b45914fa ExternalMethod optimization:
    timeit("__import__('erp5.component.extension.' + module_name, fromlist='*', level=0)",
           setup="module_name='Hoge'")
    1.754
    1.845
  * This commit:
    timeit("import_module('erp5.component.extension.' + module_name)",
           setup="from importlib import import_module; module_name='Hoge'")
    1.346
    1.311
    => Faster because import_module() calls __import__() without fromlist (not
       needed as it returns module from sys.modules after this) and level
       parameters.
    => Introduced in python3 and backported to python2.7. Recommended over
       __import__ in Python documentation for non-internal usage and when
       getting the module itself and not its top-level one.

When clicking "Reply" on a discussion post, this script adds a hidden input to mark which post we are replying to.
This cause problem  when ctrl+clicking reply to reply in a new tab, because it leaves the page with the extra hidden input, which cause problems, like clicking reply again cause `Error Value: You are not allowed to access '['1', '1']`' in this context , or clicking change page will reply.

This is a minimal effort fix, just remove the added element after click.

/reviewed-on nexedi/erp5!823

This reverts commit b7d9590c.

Users are not supposed to edit templates because documents are not in
their "normal" context, if user edit templates some unexpected things
might happen. We don't want user to change state or do some actions that
may cause interactions on other documents from a template. This was
never strongly enforced, but there was not supposed to be an easy way
for users to access the template document.

With old versions of ERP5 (before the rework of isIndexable /
isSubTreeIndexable ), users could also add lines inside the template
document and these lines where indexable.

Note that the way of editing templates is:
 * make a "real" document from the template
 * modify this document
 * use "Make Template" again - this will replace the document with the
same title.

/reviewed-on nexedi/erp5!822

This allow to pick up only few document on which to execute the action.

Listbox 'Edit' mode will contain copy/paste/delete actions.
Listbox 'Select' mode will contain include/exclude and all other erp5 actions (except delete).

Define the action on erp5 with the object_list_action category value.

Stop using change command by default.
This was a hack to hope that some parameters were propagated.

Propagate instead a cancel parameter to correctly quit dialog.

Stop hardcoding extended search parameter propagation.

This allow user to revert his change by using the browser back button.

See nexedi/erp5@9dcbc756