fix: Use individual parameters with docstring descriptions instead of Pydantic models for proper FastMCP schema generation

## Problem
The initial implementation used Pydantic models as single 'params' arguments, which caused
FastMCP to nest all parameter descriptions under the 'params' key instead of exposing them
as top-level parameters. The MCP quality scoring system expects top-level parameter descriptions
in the tools' inputSchema.

## Solution
- Changed all 8 tools from using Pydantic input models to individual parameters
- Moved parameter descriptions from Pydantic Field() to docstring Args sections
- Removed unused Pydantic input model classes (MarkdownifyInput, SmartscraperInput, etc.)
- Kept ConfigSchema as it's required for Smithery configuration

## Changes
1. markdownify: website_url as direct parameter with docstring description
2. smartscraper: user_prompt, website_url, number_of_scrolls, markdown_only as individual params
3. searchscraper: user_prompt, num_results, number_of_scrolls as individual params
4. smartcrawler_initiate: url, prompt, extraction_mode, depth, max_pages, same_domain_only as individual params
5. smartcrawler_fetch_results: request_id as direct parameter
6. scrape: website_url, render_heavy_js as individual params
7. sitemap: website_url as direct parameter
8. agentic_scrapper: url, user_prompt, output_schema, steps, ai_extraction, persistent_session, timeout_seconds as individual params

## FastMCP Schema Generation
FastMCP now generates inputSchema with top-level properties and descriptions from the
docstring Args section, which will be properly recognized by the MCP quality scoring system.

## Expected Impact
This fix should enable the +12 points for "Parameter descriptions (0/8 tools)" to be awarded,
bringing the quality score from 78/100 to 90/100 (with annotations already at +9pts).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: VinciGit00