…eeded)

Closes ScrapeGraphAI#1055

Plasmate (https://github.com/plasmate-labs/plasmate) is an open-source
Rust browser engine that outputs Structured Object Model (SOM) instead
of raw HTML. It requires no Chrome process, uses ~64MB RAM per session
vs ~300MB, and delivers 10-100x fewer tokens per page.

Changes:
- Add scrapegraphai/docloaders/plasmate.py: PlasmateLoader
  - Implements BaseLoader (lazy_load + alazy_load)
  - Calls plasmate binary via subprocess (pip install plasmate)
  - Supports output_format: 'text' (default), 'som', 'markdown', 'links'
  - Supports --selector, --header, --timeout flags
  - Optional fallback_to_chrome=True for JS-heavy SPAs
  - Async-safe: runs subprocess in executor thread pool
- Update scrapegraphai/docloaders/__init__.py: export PlasmateLoader
- Update scrapegraphai/nodes/fetch_node.py: support plasmate config dict
  in FetchNode (alongside browser_base and scrape_do)
- Add tests/test_plasmate.py: 25 unit tests (init, cmd building,
  lazy_load, alazy_load, fallback, error handling)

Usage:
  from scrapegraphai.docloaders import PlasmateLoader

  loader = PlasmateLoader(
      urls=['https://docs.python.org/3/library/json.html'],
      output_format='text',
      timeout=30,
      fallback_to_chrome=True,  # optional: retry with Chrome for SPAs
  )
  docs = loader.load()

  # Or via FetchNode config:
  graph_config = {
      'plasmate': {
          'output_format': 'text',
          'timeout': 30,
          'fallback_to_chrome': False,
      }
  }