gsl-lite: add None option for on_contract_violation

[perseoGI]

Summary

Changes to recipe: gsl-lite/all

Motivation

In this very old PR #5521, user request to have the possibility of not defining a concrete on_contract_violation policy. Instead, let downstream decide.

Details

Close #5521

---

• Read the contributing guidelines
• Checked that this PR is not a duplicate: list of PRs by recipe
• Tested locally with at least one configuration using a recent version of Conan

[AbrilRBS]

None is not a good name in this case, it gets too close to options being able to not be defined

[conan-center-bot]

Conan v1 pipeline ✔️

All green in build 1 (ff011a875c933556760e792faf360e4d9d9077b6):

• gsl-lite/0.41.0:
  All packages built successfully! (All logs)

• gsl-lite/0.40.0:
  All packages built successfully! (All logs)

• gsl-lite/0.38.1:
  All packages built successfully! (All logs)

• gsl-lite/0.39.0:
  All packages built successfully! (All logs)

• gsl-lite/0.38.0:
  All packages built successfully! (All logs)

• gsl-lite/0.37.0:
  All packages built successfully! (All logs)

---

Conan v2 pipeline ✔️

Note: Conan v2 builds are now mandatory. Please read our discussion about it.

All green in build 1 (ff011a875c933556760e792faf360e4d9d9077b6):

• gsl-lite/0.41.0:
  All packages built successfully! (All logs)

• gsl-lite/0.37.0:
  All packages built successfully! (All logs)

• gsl-lite/0.40.0:
  All packages built successfully! (All logs)

• gsl-lite/0.39.0:
  All packages built successfully! (All logs)

• gsl-lite/0.38.1:
  All packages built successfully! (All logs)

• gsl-lite/0.38.0:
  All packages built successfully! (All logs)

[jcar87]

In terms of implementation, easiest in these cases, rather than have None as a valid value, is to simply remove it from the default_options, and then access it with get_safe - it will remain undefined unless it is done explicitly. Has the same effect but it's less code (and arguably, a better reflection of the situation).

secondary to this, changing the behaviour of a default option will impact users that may already be relying on the previous default behaviour. we may break them or cause their own apps to behave differently. We need to evaluate the risk in this case (and any cases where we change a default). I would try and determine that before merging this PR.

And more generally - we should re-evalaute the pattern in Conan Center for header only libraries to propagate compiler flags to consumer via Conan options. The typical use outside of Conan would be for the consumers to do a #define before doing an #include. I would only consider the "option" approach if the upstream project is able to generate targets in the -config.cmake files where the imported targets propagate it as a compiler flag, and if this is configurable at build time. So reflect how the library may be used outside of Conan Center, rather than invent a new pattern.

[jcar87]

has this pattern not come up before, where we decide to leave the value undefined instead? I think it achieves the same effect in the package ID ?