Use write locks in license store to avoid deadlocks

[goneall]

Fixes #250

[pmonks]

I haven't run into this issue myself, so my familiarity with it is minimal, but this change looks fairly innocuous, and all of the unit tests pass for me locally (including the slow tests).

Given the potential for decreased performance in concurrent scenarios (and the fact that such aspects are difficult to test via unit tests), I'd suggest that this change be released in an alpha, beta, release candidate or similar, so that it can be tested in real world scenarios by consumers of the library (and given how large of a change v2.0.0 is, this is probably a good idea more generally anyway).

[goneall]

I'll go ahead and merge this in so it can get some early testing with the 2.0 spec updates.