Add GError arguments to e_phone_number_get_default_region() and
e_phone_number_get_country_code_for_region() to properly report
missing phone number: Warnings cannot be handled programatically
and make GTest based unit tests failed.

Ensure that libebook-contacts always gets linked using the C linker,
instead of the C++ automake would prefer, because of the optional
e-phone-number-private.cpp file. Without this change linking with
explicitly disabled phone number support (--without-phonenumber)
would fail because no C++ linker is know. Similarly the library
would depend on the C++ standard library if phone number was
disabled automatically, because libphonenumber was not found.

This is needed to properly reopen addressbooks without
specifying the full custom summary a second time.

This avoids errors of the kind "e-book-backend-file.c:1478: Failed to open
sqlitedb: table folder_id has no column named file_as" when opening an
addressbook with customized summary a second time, but with default
settings.

The sqlitedb backend and e_phone_number_compare() showed different
behavior for a few cases of national and short number matching.
This syncs the behavior.

It is not thread-safe at it changes the entire program's locale
settings. As a side effect also permit locale strings for functions
like e_phone_number_get_country_code_for_region().

e_phone_number_compare() and e_phone_number_compare_strings()
had different opinion how well "1234" and "234" match within
NANP regions. First considered them a national significant
number match as it interpreted the leading "1" as country
calling code, the later didn't. The former interpretion fits
better our E.164 normalization we do in the SQLite backend,
therefore we adopted that approach.

This also introduces e_phone_number_compare_strings_for_region().

https://bugs.freedesktop.org/show_bug.cgi?id=59571

This is mostly a matter of clarifying whose responsibility it is to
free things, in a machine-readable way.

Bug: https://bugzilla.gnome.org/show_bug.cgi?id=695151
Signed-off-by: Simon McVittie <simon.mcvittie@collabora.co.uk>
Reviewed-by: Matthew Barnes <mbarnes@redhat.com>