fix(extraction): guard field transform against None in JSON extraction

[Otis0408]

This targets the develop branch, per CONTRIBUTING.

Who hits this

Anyone using JsonCssExtractionStrategy / JsonXPathExtractionStrategy with a schema field that pairs a fail-capable extraction (a type: "regex" that may not match, or a type: "attribute" whose attribute may be absent) with an optional transform (lowercase / uppercase / strip) — all documented, sanctioned schema keys. This is a common real-world shape, e.g. a list of anchors where one <a> has no href.

Root cause

In JsonElementExtractionStrategy._extract_single_field, the transform block ran unconditionally after the type pipeline. When the pipeline yields None (a regex miss, or element.get() returning None for a missing attribute), _apply_transform calls None.lower() / None.strip() / None.upper() and raises AttributeError: 'NoneType' object has no attribute 'lower'. The very next line already treats None as a legitimate outcome (return value if value is not None else field.get("default")), so None is expected here and the unguarded transform was an oversight.

Impact depends on the path:

• baseFields — the loop in extract() has no try/except, so the AttributeError aborts the entire extraction.
• list of sub-fields — one non-matching sibling throws into _extract_field's try/except, which returns the field default for the whole list, silently dropping valid siblings.

Fix

One line — only apply the transform to a non-None value, letting None flow to the existing default handling:

-        if "transform" in field:
+        if "transform" in field and value is not None:
             value = self._apply_transform(value, field["transform"])
         return value if value is not None else field.get("default")

Before / after (public API)

from crawl4ai.extraction_strategy import JsonCssExtractionStrategy

html = """<div class='wrap'><ul>
  <li><a class='lnk' href='http://x'>a</a></li>
  <li><a class='lnk'>b</a></li>
  <li><a class='lnk' href='http://z'>c</a></li>
</ul></div>"""
schema = {
    "name": "links", "baseSelector": "div.wrap",
    "fields": [{
        "name": "links", "selector": "a.lnk", "type": "list",
        "fields": [{"name": "href", "type": "attribute",
                    "attribute": "href", "transform": "lowercase"}],
    }],
}
JsonCssExtractionStrategy(schema).extract(None, html)
# before: []  -> http://x and http://z silently lost
# after:  [{'links': [{'href': 'http://x'}, {}, {'href': 'http://z'}]}]

Tests

Adds tests/test_json_transform_none_guard.py (5 tests, CSS + XPath) covering the baseFields crash, the list-drops-valid-siblings regression, and that transform still applies to genuine values. All fail on the current code and pass with the fix.