Fix inaccuracies in Readme.md and add description of mem-mode.

[Mittagskogel]

• Use double-precision in op-mode code examples.
• Describe mem-mode with examples. Explain why only double precision works for now.

[lucasmnd]

Hi everyone, thanks for updating the documentation!

Since this PR aims to add a description of mem-mode, I wanted to ask about its current state in the codebase, since I've run into some issues. I'm currently researching mixed-precision in HPC and I'm very interested in using the mem-mode to generate the error heatmap described in Section 6.3 of the paper.

Will this PR also address how to properly build and invoke this error sensitivity tracking? Is the heatmap feature functional in this current branch, or is it maintained somewhere else? Any guidance would be incredibly helpful for our research!

[Mittagskogel]

Hi @lucasmnd, thank you for reaching out! Take a look here: #18 (comment)

[MinliLiao]

The C++ part of mem-mode is missing the information on declaring the __raptor_truncate_mem_func, __raptor_truncate_mem_value and __raptor_expand_mem_value in text.
The C++ code example has errors: 1) truncation of b is not stored back in b. 2) there's an extra ); for the truncation of foo.

I'm not familiar with fortran code, but it doesn't seem to have the same problem, so I assume it's good?