Add test to show that changing option values in config_options is a no-go

[AbrilRBS]

Changelog: Omit
Docs: Omit

Probably won't merge it, just wanted to be able to show it and have a place to link from CCI

cc @uilianries - as part of my review for https://github.com/conan-io/conan-center-index/pull/24627/files#diff-a4348270915a235071ac4a599f4c2cc7d2427bb96444198c014796e55edff14eR54

[uilianries]

Is it expected, right? It's same behavior since Conan 1.x.

[jcar87]

This is how config_options has always worked - if an option is given a value (conditional on something), it will take that value "as default" as if it had been defined in the default_options dictionary. Much like the default options dictionary, if the user provides a value in the CLI, then that's the value that taken - it makes sense: the user gave us a value, so we use it.

It can be useful to have a default option value conditional on a setting - it can't be conditional on other options, because the options dont have a value (yet).

What is less clear is hat the behaviour is when an option has a default_option and is also conditionally set in config_options - not sure that bit was ever documented.

But changing this behaviour is likely to break things.

[memsharded]

This was different in Conan 1.

The self.options.foo = 2 had higher priority than the profile value in Conan 1