fix(e6): escape lone apostrophes in literals under FIX_QUOTE_ESCAPES

[tkaunlaky-e6]

Summary

Kantar's INSERT queries containing Databricks double-quoted strings with an apostrophe inside — e.g. "SCARLET'S WALK (2023 REMASTER)" — were emitting invalid E6 with the apostrophe unescaped. The bare ' terminated the output string literal early.

The E6 generator's escape_str override at sqlglot/dialects/e6.py unconditionally skipped the single-quote escape pass under FIX_QUOTE_ESCAPES=true. That is correct only for Variant C literals (parser-merged adjacent string tokens with '' embedded in Literal.this). For literals whose text contains a lone ' (Databricks "..." strings, or \' escapes the tokenizer consumed), the override produced raw apostrophes in output.

Fix

Gate the override on whether '' is actually present in the literal text. When '' is not in the text, defer to the base Generator.escape_str, which correctly replaces ' with \' since E6's STRING_ESCAPES = ['\\']. When '' is present, retain the existing behavior — '' passes through unchanged, no \' introduced.

if os.getenv("FIX_QUOTE_ESCAPES", "False").lower() == "true":
    if "''" not in text:
        return super().escape_str(text, escape_backslash)
    # ... existing override body

Behavior

Input	Literal.this	Path	Output
"SCARLET'S WALK" (Kantar)	SCARLET'S WALK	base	'SCARLET\'S WALK'
'ROCKIN'' AROUND'	ROCKIN'' AROUND	override (unchanged)	'ROCKIN'' AROUND'
'Côte d''''Azur'	Côte d''''Azur	override (unchanged)	'Côte d''''Azur'

Test plan

• Existing test_fix_quote_escapes continues to pass (4 cases — all literals contain '', take the unchanged path).
• Two new cases added: Databricks double-quoted string with apostrophe inside, and the full Kantar INSERT shape.
• Full E6 dialect test suite: 50/50 pass.
• Verified end-to-end against the running converter API with FIX_QUOTE_ESCAPES=true on Kantar's exact query.