[GEN][ZH] Fix several potential buffer overruns in game code

[slurmlord]

This PR addresses some buffer size issues that had the potential of writing beyond the bounds.

• The wide char methods FormatMessageW, GetDateFormatW and GetDateFormatW take the buffer size in (wide) characters, not bytes.
• The copying of colors in dynamesh.h seem like a simple bug, as the correct behavior is just below in the same method.
• In the case malloc was called in FastAllocator, assert that it wasn't 0 before dereferencing it. Not a buffer overflow but still better with an assertion than an access violation, IMO.
• For ftp, the buffer is being used in Cftp::GetNextFileBlock for a strncpy call with a hard-coded limit of 256.
• The change in rinfo is to compare with 1 below the array size, as OverrideFlagLevel is immediately incremented before indexing into the array.

[xezon]

The double loop is not the correct location to write to this member.

[slurmlord]

Agreed, it was a compromise in order to keep the diff as minimal as possible. Went with a separate loop for clearing m_current_D3DSurface instead.

[xezon]

Does any of these changes fix an observable bug in the game?

[slurmlord]

Does any of these changes fix an observable bug in the game?

Not that I know of, but I'm unfortunately not familiar enough with the set of known bugs either to tell. The issues were identified with the MSVC static code analysis and some manual check to see if it seemed plausible or not, so it's approaching from a bit of a different angle than a known repro.