MCP compliance, safety and math fixes; English tool docs; symbolic dsolve

Fork of https://github.com/thinkitpossible/CalcMCP with:
- tools/list now publishes inputSchema (upstream used non-standard 'params'
  key, leaving clients with no parameter information)
- tool descriptions and parameter docs rewritten in English with examples
- JSON-RPC notifications are no longer answered; tool failures set isError
- raw eval() of tool arguments removed (json.loads/ast.literal_eval for
  matrices, sympy.lambdify for the numeric ODE right-hand side)
- complex eigenvalues serialize as 'a+bi' strings instead of crashing
- solve accepts '=' equations; ttest accepts {"popmean": ...} params
- new symbolic_tool operation 'dsolve' for symbolic ODE solutions

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

README.md

@@ -1,9 +1,26 @@
 # Scientific Calculator MCP Server
 
-> A production-ready Model Context Protocol (MCP) server providing advanced mathematical calculation capabilities for AI models. Supports symbolic math (SymPy), numerical computing (NumPy/SciPy), data analysis (pandas), and image processing.
+> A Model Context Protocol (MCP) server providing advanced mathematical calculation capabilities for AI models. Supports symbolic math (SymPy), numerical computing (NumPy/SciPy), data analysis (pandas), and image processing.
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 
+> **Fork notice** — this is a patched fork of
+> [thinkitpossible/CalcMCP](https://github.com/thinkitpossible/CalcMCP).
+> Changes relative to upstream:
+>
+> - **MCP compliance**: tools are published with a proper `inputSchema` (upstream used a
+>   non-standard `params` key, so clients saw no parameters at all); JSON-RPC notifications
+>   (`notifications/initialized`) are no longer answered with an error; tool failures set
+>   `isError: true` on the result.
+> - **Tool descriptions rewritten in English**, with per-parameter docs and inline examples.
+> - **Security**: removed all raw `eval()` of tool arguments (matrix inputs now parse via
+>   `json.loads`/`ast.literal_eval`, the numeric-ODE right-hand side via `sympy.lambdify`).
+> - **Bug fixes**: eigenvalue/eigenvector results with complex values no longer crash JSON
+>   serialization (rendered as `a+bi` strings); `solve` accepts equations written with `=`;
+>   one-sample t-test accepts `params` as `{"popmean": ...}` or a bare number.
+> - **New**: `symbolic_tool` operation `dsolve` — symbolic ODE solving
+>   (e.g. `Derivative(y(x), x, 2) + y(x) = 0` → `y(x) = C1*sin(x) + C2*cos(x)`).
+
 ## Quick Start
 
 ### 1. Install Dependencies