Clean build warnings

[egorpugin]

Currently we see tons of warnings during MSVC builds.
They should be fixed during modernizing code phase.
See this log for warnings https://ci.appveyor.com/api/buildjobs/3v0vmv6xjy658f79/log

[stweil]

Similar warnings can be produced on Linux using clang++. But even the default debug build with gcc produces tons of warnings (mainly signed/unsigned mismatch).

[egorpugin]

I've also removed /W1 (warning level 1, default is 3) in leptonica for MSVC.
So, @DanBloomberg probably should look at those warnings in lept too.
https://ci.appveyor.com/api/buildjobs/0s1o3vud8x1e1pe6/log

[DanBloomberg]

In leptonica:

• 95% of these are because I use floats, and floating point numbers (like 1.0) are assumed to be doubles.
• Most of the rest are implicit conversions from float to int, where the ints are indices into a 2d image array.

[egorpugin]

• 1.0 -> 1.0f
• make casts explicit

What do you think?

[stweil]

Mixing float and double can result in additional code (and execution time) for conversions between both types.

[DanBloomberg]

Any program may do a few tens of these, so we're talking about a small fraction of a microsecond.

Eliminating these windows compiler warnings is not high on my todo list. I'll try to remember to add "f" to numbers in files that I'm modifying for other reasons.

[zdenop]

Closing this, because more info is provided in #1036