added authentications capabilities for http checks

See merge request arn/picomon!12

picomon/checks.py: fix missing parenthesis in CheckTLSCert().

See merge request arn/picomon!11

Adding new check: CheckTLSCert{4,6} for testing x509 certificate's expiration over TLS connection.

See merge request !10

Add support for dynamic DNS resolution in host addresses



See merge request !9

Change binary (and check definition) for check OpenVPN.



See merge request !8

Don't keep In-Reply-To between success/failure cycles not to have all
reports from a particular check in the same thread.

Fixes a bug introduced in the previous commit,
3c3ca1cf

Reloading of the configuration implemented. Old/changed checks in failure send
an email saying it will never come back again. New checks are added to the list.
Unmodified checks are kept the same, preserving context.
Systemd unit now knows about reload and by the way uses systemd itself as a
watchdog instead of the one provided alongside the program.

This removes some useless classes/__repr__/redefinitions and
cleans-up the check's destination handling between classes and in emails.

With systemd it's easy to manage several instances of picomon (via templated
units) so this should be quite usefull.
Also we switch to /etc/picomon for config (instead of %prefix%/etc/picomon)
as /usr/local/etc don't seem to exist for real and that way it's not abnormal
to ship the unit in /etc too. Also it's easier to manage the config at only
one place (think about etckeeper etc.).

This fixes a strange "Error on receive" issue on some check_http versions.
Enabling SNI we now also check the connection using the certificate of the
configured vhost (and not the default one).

We used to drop any mail we tried to send, regardless of whether
sending it actually succeeded or not.  This could result in lost mails
if the server failed to accept some mails for any temporary reason.

Fix this by re-queuing emails that failed to send and wait a few
moments before retrying.

The code was moved recently to be accessible by the watchdog but
a variable name was not changed…

Prints redirected to files are heavily cached, so we need to flush the
stream if we want to be able to read the output in log files as soon as
we return from signal handler.

Since we never reclaim finished tasks, exceptions raised during thread
run are never seen. Add a runner call in between to display them.

As the user configuration is now loaded after the mails module, there
is a race in the mail thread that fetches configuration options, which
may or may not already have the user's value.

So, remove caching of those value, which leads for them to be read much
later after everything has properly been initialized.  The timeout
setting might still have the default value on the first loop run, but
this doesn't matter as the server isn't started at this point anyway.