Fix bugs in dijkstra algorithm and csv parsing

[reki2000]

Fixes three bugs and one refactoring

1. Dijkstra loop should check if the popped (from priority queue) distance for node i is smaller than calculated distance d[i]. This r reduces the complexity of main the loopto O( (|V|+|E| log |V|) - don't know the original's complexity - Thanks TsumoiYorozu !

2. Fill distance array d in Dijkstra loop with enough big value like INT32_MAX to reduce d[i] == 0 (checks if the node has been already visited)

3. CSV parsing logic has a critical bug - parses integer without decimal dot to x1000 instead of x100.

4. Use Int32 for Distance type explicitly. But still not completed over all languages in this benchmark.

[wx257osn2]

CSV parsing logic has a critical bug - parses integer without decimal dot to x1000 instead of x100.

😱 I apologize for that...

[reki2000]

@wx257osn2 いえ、元のコードからあったバグで、修正いただいたおかげて発見できました！