feat: implement XER MCP Server with 9 schedule query tools

Implement complete MCP server for parsing Primavera P6 XER files and
exposing schedule data through MCP tools. All 4 user stories complete.

Tools implemented:
- load_xer: Parse XER files into SQLite database
- list_activities: Query activities with pagination and filtering
- get_activity: Get activity details by ID
- list_relationships: Query activity dependencies
- get_predecessors/get_successors: Query activity relationships
- get_project_summary: Project overview with counts
- list_milestones: Query milestone activities
- get_critical_path: Query driving path activities

Features:
- Tab-delimited XER format parsing with pluggable table handlers
- In-memory SQLite database for fast queries
- Pagination with 100-item default limit
- Multi-project file support with project selection
- ISO8601 date formatting
- NO_FILE_LOADED error handling for all query tools

Test coverage: 81 tests (contract, integration, unit)

src/xer_mcp/parser/table_handlers/base.py

@@ -0,0 +1,30 @@
+"""Base class for XER table handlers."""
+
+from abc import ABC, abstractmethod
+
+
+class TableHandler(ABC):
+    """Abstract base class for XER table handlers.
+
+    Each handler is responsible for parsing a specific table type
+    from the XER file and returning structured data.
+    """
+
+    @property
+    @abstractmethod
+    def table_name(self) -> str:
+        """Return the XER table name this handler processes (e.g., 'PROJECT', 'TASK')."""
+        ...
+
+    @abstractmethod
+    def parse_row(self, fields: list[str], values: list[str]) -> dict | None:
+        """Parse a single row of data from the XER file.
+
+        Args:
+            fields: List of column names from the %F line
+            values: List of values from the %R line
+
+        Returns:
+            Dictionary of parsed data, or None if the row should be skipped
+        """
+        ...