fix(filters): separate each match on its own line for source: watches

[Yi-111-a]

Fixes #4477

The problem

A watch on a source: URL passes append_pretty_line_formatting=False to the filter helpers, and its filtered text is then kept verbatim instead of being run through html_to_text():

# changedetectionio/processors/text_json_diff/processor.py
append_pretty_line_formatting=not self.watch.is_source_type_url,
...
if watch.is_source_type_url:
    # For source URLs, keep raw content
    stripped_text = html_content

TEXT_FILTER_LIST_LINE_SUFFIX is "<br>", which only works as a line separator because Inscriptis turns it back into a newline later on. For a source: watch that step never runs, so the code skipped the separator entirely — and with no separator at all, every match ran straight into the next on one line.

Reported on source:https://www.wikipedia.org/ with the //link/@href filter:

Current:
/static/apple-touch/wikipedia.png/static/favicon/wikipedia.ico//creativecommons.org/licenses/by-sa/4.0///upload.wikimedia.orghttps://wikis.world/@wikipedia

The fix

filter_match_separator() picks a real "\n" instead of "<br>" when the result will not be passed through Inscriptis:

if not append_pretty_line_formatting:
    return "\n" if len(html_block) else ""

xpath_filter(), xpath1_filter() and include_filters() all carried the same if append_pretty_line_formatting and len(html_block) and ... condition, so all three now call the helper. That also fixes the CSS and xpath1: variants of the same bug, which had the identical symptom.

Two details preserved:

• The first match still never gets a separator, so no leading newline is introduced.
• The Inscriptis path is unchanged, byte for byte — same <br> marker, same skip for br/hr/div/p (which already break a line).

Result for the reported case:

/static/apple-touch/wikipedia.png
/static/favicon/wikipedia.ico
//creativecommons.org/licenses/by-sa/4.0/
//upload.wikimedia.org
https://wikis.world/@wikipedia

Tests

New changedetectionio/tests/unit/test_filter_match_separator.py covers the multi-match case for XPath, xpath1: and CSS, the single-match and no-match cases (no stray leading newline), and the helper itself for both paths.

changedetectionio/tests/unit/ + test_xpath_selector_unit.py   467 passed
test_css_selector, test_xpath_selector, test_xpath_default_namespace,
test_source, test_lxml_concurrency, test_filter_exist_changes,
test_ignore_text                                             44 passed