Improve LimitingTunables implementation (Tunables example)

[webmaster128]

Description

1. Rename the configured memory limit to limit consistently to avoid confusion with maximum.
2. Adjust memory type using the limit in memory_style. Without this change, memories that have no maximum when requested derive the style from the requested memory, not from the adjusted one. So when running this example on Windows, we got a dynamic memory before this PR and get a static memory now.
3. Since the memory_style signature does not allow errors, the logic was changed to an adjust memory to something potentially broken, validate later approach. Do we really want to keep this limitation? Or should implementors be allowed to return errors in memory_style?

Ps.: I'm heavily unit testing this in our project. But I assume those don't belong here, right?

Review

• Add a short description of the the change to the CHANGELOG.md file

[syrusakbary]

bors r+

[bors]

Build succeeded:

• Audit
• Code lint
• Test on linux
• Test on linux-aarch64
• Test on macos
• Test on windows