intrinsics module with alternative implementations

[jalvesz]

Add intrinsics module containing replacements for intrinsic function where some feature is found interesting: faster implementation, better accuracy, both simultaneously.

This PR follows the discussion in discourse https://fortran-lang.discourse.group/t/lfortran-now-supports-all-intrinsic-functions/8844/41 and it's based on https://github.com/jalvesz/fast_math

• sum: 2 options (stdlib_sum and stdlib_sum_kahan)
• dot_product: 2 options (stdlib_dot_product and stdlib_dot_product_kahan)

cc: @fortran-lang/stdlib @perazz @certik @jvdp1

[jalvesz]

One philosophical question: should the fsum interface be renamed to sum to enable direct replacement of the intrinsic? Keep this name? Or yet something like stdlib_sum? (Same for fprod->dot_product)

Regarding the kahan versions, given that the accuracy gains are close between the pure chunked version and the kahan one, I'm wondering which level of support should be enabled to switch between them?

[perazz]

I prefer to keep stdlib_sum and stdlib_dot_product

LGTM @jvdp1 @jalvesz!

[perazz]

LGTM pending a few comments, thank you @jalvesz I think it is almost ready! I also like @jvdp1's suggestion of adding integer support.

[perazz]

LGTM, thank you @jalvesz. I've adjusted the description of the kahan versions, that do not support integer in the final revision. Great addition!

[jalvesz]

Thanks @perazz I guess I pushed the changes before checking online that you had made the changes as well 😅 the same modifications are ready in the last commit

[jvdp1]

LGTM. It can be merged pending a few very minor changes.