Fix shadowcasting for weather with sight_penalty != 1.0

[SurFlurer]

Summary

none

Purpose of change

Fix the comment #65205 (comment)
Fix #72508
Partially address #65205 ( there are still inconsistent shadowing in different weather conditions, but no more see through the wall )

If the span split is due to two transparent map tiles having different transparency ( like between indoor tiles and outdoor tiles ), those squares will be checked by both B and C spans. I could't say I understand shadowcasting very well, but I believe this behavior led to performance hit and weird shadowing in the steel mill.

Describe the solution

Like skip_first_row, add a variable skip_first_column in struct span, which indicates the span should skip the squares whose leading_edge_minor is the same as the span's start_minor. The attribute is to be inherited by A, B and D spans ( those spans have the same start_minor as the span being split. ) As for C span, after being split, the attribute is rechecked and reapplied, to avoid inappropriately skipping squares.

Describe alternatives you've considered

Testing

There is no perceived performance hit in the steel mill, even if I turned off compiler optimization for shadowcasting code.

There's more correct shadowing in the steel mill: ( compared with #65205 (comment) )

For #65205, inconsistency still exists, but the wall correctly blocks the terrain behind it.

Generally, this is not perfect nor final solution to artifacts in shadowcasting, but I believe I'm making it more correct.

Additional context