feat(mo): add moRealVectorNeighbor for continuous local search

[evvaletov]

Summary

Adds moRealVectorNeighbor (mo/src/neighborhood/): a neighbor type for
real-valued vector solutions with delta-based forward/backward moves via
moBackableNeighbor. Follows the same <EOT, Fitness> template convention
as moIndexNeighbor.

Motivation

Paradiseo's MO module has neighbor types for index-based and bit-based
representations but nothing for real-valued vectors.
moRealVectorNeighbor is needed by the real-valued neighborhood explorer
in a follow-up PR (#87).

Changes

File	Type
mo/src/neighborhood/moRealVectorNeighbor.h	New

Additive only — no existing files modified.

[nojhan]

Hi! Thanks for the contributions, it's really welcomed!

However, I cannot do a code review, because the PR contains almost all Paradiseo history, which completely breaks Github's review feature. Can you make a new branch from HEAD, cherry-pick your contributions in two commits and then force push on the PR branch?

[nojhan]

From the content description, here is what I can say:

• The continuous neighborhood is a cool feature which was missing.
• I think the continuator can be achieve by composing other operators.

About this last remark: Paradiseo is designed so that you can always assemble atomic operators to make one with broader feature set. Here, you want to make a continuator (or maybe just a monitor? I'm not sure) with two features. The right approach would thus be to combine the operator counting seconds with the one counting generations, not to have a new one that combine the two. Once understood, you will be able to combine even more continuators/monitors... That's the beauty of Paradiseo :-)

[evvaletov]

You're right, the continuator is a composition problem, not a new class. I've removed eoSecondsElapsedTrackGenContinue from this PR. The generation count is already available through eoIncrementorParam wired into the checkpoint, exactly as make_checkpoint.h demonstrates.

The PR now contains only moRealVectorNeighbor, which you mentioned was a missing piece.

I've also rebased onto HEAD so the diff is reviewable. Sorry about the history mess: the fork had a divergent merge commit from 2013 that gave every subsequent commit a different SHA.

[nojhan]

Awesome. Only very minor comments, the code is good.

[nojhan]

This header is at least wrong on the dates. And there should be only one copyright mention.

[nojhan]

Website and contact are also wrong (maybe this will be fixed by PR #88 ?).

[nojhan]

Needs an empty line after the summary line.