lxml: drop self.uistate.window from module-level ErrorDialog calls

`lxml/lxmlGramplet.py:72,94` reference `self.uistate.window` at
module top level — both inside `except ImportError:` branches that
guard `import gzip` and `from lxml import etree, objectify`. `self`
is undefined at module-load time (the gramplet instance does not
exist yet), so when either import actually failed the addon module
crashed with:

    NameError: name 'self' is not defined

…instead of showing the intended ErrorDialog. Ruff (F821) flags
both:

    F821 Undefined name `self`
       --> lxml/lxmlGramplet.py:72:69
       --> lxml/lxmlGramplet.py:94:104

PR gramps-project#820's body listed this as one of the "Real bugs":
"stray `parent=self.uistate.window` at module level in
lxmlGramplet".

Drop `parent=self.uistate.window` from both `ErrorDialog` calls.
The dialogs render with no parent window — not ideal aesthetically,
but at module-load there genuinely is no UI parent to anchor to (the
alternative — moving these checks into the gramplet's `__init__` —
is a larger restructuring outside this PR's scope).

In practice the gzip branch is dead code (gzip is in the Python
stdlib and never raises ImportError); the lxml branch is the
user-visible win — when a user installs the addon without the lxml
package present, they now get the intended "Missing python3 lxml"
dialog instead of an opaque NameError that masks the underlying
cause.

No regression test is added for this fix:

  - The bug only fires when `from lxml import ...` raises
    ImportError, which requires either an unusual install or
    subprocess-isolated sys.modules tampering to simulate.
  - The natural location for the test would be `lxml/tests/`, but
    the addon directory name (`lxml`) collides with the third-party
    `lxml` package — Python's import system treats the latter
    (regular package, /usr/lib/python3/dist-packages/lxml) as taking
    precedence over the namespace-package addon directory, so
    `python3 -m unittest lxml.tests.test_*` cannot reach the
    addon's tests subtree on any system that has `lxml` installed
    (which the CI image always does).
  - Ruff F821 catches future regressions of this exact pattern,
    which is the same gate that surfaced the bug in the first
    place.

Verified via `ruff check --select=E9,F63,F7,F82 --no-fix
--exclude='*.gpr.py' lxml/` (both F821 sites on `self` removed —
the remaining F632 sites at lines 400/758 are in-flight via PR gramps-project#837)
and `python3 -m py_compile lxml/lxmlGramplet.py`.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: 2 people

lxml/lxmlGramplet.py

@@ -69,7 +69,7 @@
     GZIP_OK = True
 except ImportError:
     GZIP_OK = False
-    ErrorDialog(_('Where is gzip?'), _('"gzip" is missing'), parent=self.uistate.window)
+    ErrorDialog(_('Where is gzip?'), _('"gzip" is missing'))
     LOG.error('No gzip')
 
 #-------------------------------------------------------------------------
@@ -91,7 +91,7 @@
     LIBXSLT_VERSION = etree.LIBXSLT_VERSION
 except ImportError:
     LXML_OK = False
-    ErrorDialog(_('Missing python3 lxml'), _('Please, try to install "python3 lxml" package.'), parent=self.uistate.window)
+    ErrorDialog(_('Missing python3 lxml'), _('Please, try to install "python3 lxml" package.'))
     LOG.debug('No lxml')
 
 #-------------------------------------------------------------------------