UTF-8 encoding fixes + tests for xcp.pci, .cmd and .net.biosdevname

[bernhardkaindl]

• The changes are unit-tested using the included testsuite improvements to test UTF-8 handling.
• The changes to xcp.pci.PCIIds and xcp.net.biosdevname are tested by @alexhimmel using XenRT with his Python3 port of the ACK xapi-plugin.

The commits of 1-21 by ydirson are from #17, this is just to allow for a separate review of my additional commits.

Foundational information:

For reading UTF-8 characters from files using Python3, there AFAIK are 3 possibilities:

1. On Unix, the Python3 interpreter's Locale must be set to an UTF-8 locale
2. Files are opened using open(file, encoding="utf-8"), subprocess pipes are opened using Popen(..., encoding="utf-8")
3. Binary mode/Bytes are used for I/O (open(file, "b") and then all data read and written is encoded and decoded using explicit encode()/decode() alls using the utf-8 codec.
   is needed.

My observations on these are:

1. While XenServer has en_US.UTF-8 in /etc/locale which shell logins get passed as LANG=, it may not always be set. For example, when daemons are started it is good practice to clear the environment. The ACK xapi-plugin appears be an example for this situation. The locale can be enforced interpreter-wide by setting LC_CYTPE/LC_ALL using the environment or locale.setlocale(), but it affects the entire process including all threads, calls to setlocale are not thread-safe and it would not be good practice for a library like python-libs/xcp to change the locale used by the process.
2. encoding='utf-8' works nice, is easy to apply to all open and Popen calls, but is neither supported nor needed for Python2. Passing to all open and Popen can be done thru a **kwargs keyword parameter which passes the needed arguments as a dict.
3. Adding encode and decode calls to all locations where data would need to be encoded from string to bytes and from bytes to string, and ensuring that it is not over-done can go wrong. For example, a encode(args) call on bytes internally results in decode().encode(args), which can go wrong and can even result in raised UnicodeDecodeError because the implicit decode() is done without encoding= and errors= arguments.

Option 2, passing encoding="utf-8", errors="replace" (replaces decoding errors with "?", without raising an exception) using a **kwargs keyword parameter dict on python3 and empty on python2 is the easiest and safest option that I can imagine.

[psafont]

The DCO check is failing, this is because the signoffs are missing from some commits. To easily sign them of, run git rebase --signoff 09ba68aca491f26e8552e935fe36a6af634b55a5

[bernhardkaindl]

All commits of this PR have now been refactored and merged/obsoelted by PRs which are now merged.