Avoid repeated global strings

[LucasSte]

When debugging the issue #1367, I found out that we keep emitting repeated global strings. The LLVM IR for the contract pointed in such an issue had the string math overflow,\0A defined 1008 times.

This PR DOES NOT solve #1367, but it is something that needed a fix.

[xermicus]

Are these "global strings" guaranteed to be readonly?

[seanyoung]

Are these "global strings" guaranteed to be readonly?

only when constant is set

[LucasSte]

Are these "global strings" guaranteed to be readonly?

only when constant is set

I missed that. I should reuse them when they are constant. Let me update the PR.

[LucasSte]

Are these "global strings" guaranteed to be readonly?

only when constant is set

Fixed!
The changes worked because presently we do not generate any non-constant global string.

[seanyoung]

How are we testing this?

[LucasSte]

How are we testing this?

The test file I added tests/codegen_testcases/solidity/global_strings.sol generates the LLVM-IR file and reads it. It ensures that we only have a single math overflow string there, instead of 1008.

[xermicus]

Does the contract need all those empty functions for this case?

[LucasSte]

I can shorten the contract, but I left it as it is because it used to generate 1008 definitions of math overflow.
What is your opinion in shortening it?

[xermicus]

Well, it's not the end of the world. I'll leave that up to you. I just thought that if in the end only 1 single math overflow string instead of 1008 should be generated, the test code could be more to the point.