Decode string context straight to using StorePaths

[Ericson2314]

I gather decoding happens on demand, so I hope don't think this should
have any perf implications one way or the other.

[thufschmitt]

The last commit looks good to me. Apart from the potential conflicts, is there any reason for this to depend on #6236 ?

[Ericson2314]

@thufschmitt It is probably not necessary, but I was a bit nervous about making a Store more accessible to the eval cache, because I didn't want to risk some any bit of store state introducing impurities.

[Ericson2314]

Squashed out the second commit so there is is no StoreDirConfig config in the history. (I will make it part of that PR.)

[dpulls]

🎉 All dependencies have been resolved !

[Ericson2314]

@edolstra how is this? The core change is the switch to StorePath.

The auxillary change is new names for strong contexts, but I think this is good:

1. Matches SearchPath and SearchPathElem.

2. Allows changes like the Path to StorePath to be done easily, "catch" all the std::pairs that need to be so changed via variable substitution. For RFC 92 (Dynamic derivations RFC 92 #4628) I will need to again modify NixStringContextElem, so this isn't even that hypothetical a benefit after this one specific instance of changing the definition.

[edolstra]

Thanks!

[Ericson2314]

Thank you!