eeschema connectivity - Adjust performance in init

This allow single init for map structures, improving first-run
performance.

eeschema/connection_graph.cpp

@@ -653,7 +653,7 @@ void CONNECTION_GRAPH::buildConnectionGraph()
 
     // We don't want to spin up a new thread for fewer than 8 nets (overhead costs)
     size_t parallelThreadCount = std::min<size_t>( std::thread::hardware_concurrency(),
-            ( m_subgraphs.size() + 7 ) / 8 );
+            ( m_subgraphs.size() + 3 ) / 4 );
 
     std::atomic<size_t> nextSubgraph( 0 );
     std::vector<std::future<size_t>> returns( parallelThreadCount );

eeschema/sch_item_struct.cpp

@@ -81,7 +81,7 @@ void SCH_ITEM::ViewGetLayers( int aLayers[], int& aCount ) const
 
 bool SCH_ITEM::IsConnected( const wxPoint& aPosition ) const
 {
-    if( m_Flags & STRUCT_DELETED || m_Flags & SKIP_STRUCT )
+    if( ( m_Flags & STRUCT_DELETED ) || ( m_Flags & SKIP_STRUCT ) )
         return false;
 
     return doIsConnected( aPosition );
@@ -90,18 +90,14 @@ bool SCH_ITEM::IsConnected( const wxPoint& aPosition ) const
 
 SCH_CONNECTION* SCH_ITEM::Connection( const SCH_SHEET_PATH& aSheet ) const
 {
-    SCH_CONNECTION* conn = nullptr;
-
-    if( m_connection_map.count( aSheet ) )
+    try
     {
-        conn = m_connection_map.at( aSheet );
+        return m_connection_map.at( aSheet );
     }
-    else
+    catch( ... )
     {
-        // TODO(JE) should we just call InitializeConnection here?
+        return nullptr;
     }
-
-    return conn;
 }