feat: add preserve_classes/preserve_tags whitelist to PruningContentFilter (#1900)

[hafezparast]

Summary

Adds preserve_classes and preserve_tags params to PruningContentFilter so users can protect specific elements (author names, timestamps, attribution) from being pruned.

Addresses #1900

The Problem

md-fit uses density-based scoring to strip boilerplate. Short metadata elements (usernames, bylines, timestamps) score low and get removed alongside actual boilerplate — losing "who said what" on discussion pages.

The Fix

Two new optional params on PruningContentFilter:

PruningContentFilter(
    preserve_classes=["author", "byline", "comment-header"],
    preserve_tags=["time", "cite"],
)

Whitelisted nodes skip scoring entirely — always kept. Default is empty sets, so existing behavior is unchanged.

Usage

from crawl4ai import AsyncWebCrawler, CrawlerRunConfig, DefaultMarkdownGenerator
from crawl4ai.content_filter_strategy import PruningContentFilter

config = CrawlerRunConfig(
    markdown_generator=DefaultMarkdownGenerator(
        content_filter=PruningContentFilter(
            preserve_classes=["author", "byline", "username"],
            preserve_tags=["time", "cite"],
        )
    )
)

async with AsyncWebCrawler() as crawler:
    result = await crawler.arun(url="https://gist.github.com/...", config=config)
    print(result.markdown.fit_markdown)  # Now includes author names

Test plan

• 20 unit tests passing
• Existing pipeline tests pass (16/16)
• Default behavior unchanged (empty whitelist)
• Nav/footer still removed (excluded_tags runs before pruning)
• Nonexistent classes/tags in whitelist are harmless

🤖 Generated with Claude Code