This commit upgrades our PDF/A compliance from 3B to 3A, and fixes a few issues previously undetected due to the different PyPDF libraries we're currently supporting that made the previous PDF fails even the 3B validation.

Improvement 1:

PDFA validators were previously detecting our file as 3B. Hence we update the metadata content `conformance` to `A`.

Issue 2:

When using `._pypdf` library, we failed the 6.1.2-2 and 6.8-1 rule even
though we have implemented them on the previous version. It seems that
this is caused by the `if SUBMOD...` check only ensuring it's not equal
to `_pypdf2_2` (which makes it trigger for the new `_pypdf`). Hence, we
reclarify the comments and fix the IF check.


Issue 3:

After implementing issue 2, it seems that a traceback occurs every time we're using `._pypdf` and calling the pdf write method. This is because the added characters on the header can't be decoded with `UTF-8`. Hence we change it to other greater-than-127-bytes characters that can still be decoded with `UTF-8`. (The actual character used here doesn't matter).

Improvement 4:

To be compliant with the new 3A rules (additional rules not there in 3B when we first implemented them), we add a minimal mark info dictionary and document structure on the PDF catalog object (`_root_object`).

task-None