GOPROXY is used to download packages from fastest source,
disable it will let all packages download from github.com

 Going from 1.17.6 to 1.17.13 gives us security and bug fixes.
 See at official changelog for details:
    https://go.dev/doc/devel/release#go1.17.minor

Passed tests: https://erp5js.nexedi.net/#/test_result_module/20221219-8E2857F

/reviewed at nexedi/slapos!1304
/reviewed-by @kirr

This is a fixup of 431eed50

The version of pandas was overwritten in
02fad6a8 but didn't work with python2.

Then the version was downgraded to 0.24.2 in
15569ea9 but ERP5 doesn't work with this
pandas version:
Traceback (most recent call last):
  File "/srv/slapgrid/slappart3/t/dgg/soft/63c49071f5be63057dd48d7a85c10226/bin/runTestSuite", line 325, in <module>
    import Products.ERP5Type.tests.runTestSuite
  File "/srv/slapgrid/slappart3/t/dgg/soft/63c49071f5be63057dd48d7a85c10226/parts/erp5/product/ERP5Type/__init__.py", line 92, in <module>
    from . import ZopePatch
  File "/srv/slapgrid/slappart3/t/dgg/soft/63c49071f5be63057dd48d7a85c10226/parts/erp5/product/ERP5Type/ZopePatch.py", line 30, in <module>
    from Products.ERP5Type.patches import Restricted
  File "/srv/slapgrid/slappart3/t/dgg/soft/63c49071f5be63057dd48d7a85c10226/parts/erp5/product/ERP5Type/patches/Restricted.py", line 491, in <module>
    allow_type(pd.indexes.range.RangeIndex)
AttributeError: 'module' object has no attribute 'indexes'

So reuse old version 0.19.2

version 0.25.3 is not python2 compatible.

Use version 0.24.2 as it is compatible with both python2 and python3.7
(even though we have no reason to use python3.7 nowadays).
See https://pypi.org/project/pandas/0.24.2/

This is a followup of 0d717989

This is a followup of 02fad6a8

This is a frontend used to provide information about various aspect of the
master instance status, like rejected slaves, failed failsafe slave
publication and maybe more.

Due to various reasons there might be an error while publishing information
about a slave, and such error would end up with stopping the processing of
the partition. This is undesired effect, thus switch to publish_failsafe style
in order to continue processing.

Publish only for active slaves.

Empty extra_slave_instance_list is a leftover of never fully working approach
of slaveless caddy-frontend implementation, so just drop it.

See merge request nexedi/slapos!1280

In some contexts it's required to not stop buildout processing on publishing
errors, thus such recipe is required in those cases.

By providing -error-status-file it's possible to check status of the publish,
the file will be created on error and removed on successful call or during
uninstall. It's responsibility of the caller to handle such errors, for example
by using proper promises.