Honor dependencies when applying the series

[bjoto]

Currently, dependencies to series (e.g. tagged with prerequisite-* are not honored, leading to that the series will probably fail to apply.

Rationale (from #riscv)

<dlan> for who running the linux-riscv kernel ci, it probably also need to
       handle the patch dependency, parse those prerequisite-* info     [16:21]
<palmer> that's @bjoto and Alex (don't think he's in here)	        [18:18]
<bjoto> dlan: The patchwork CI? It's always a "what should we test" for a CI.
	The patchwork CI has the stance of "does the patch applies cleanly for
	the maintainer branches", so it's not going at any lengths to be smart
	about applying patches.					        [08:28]
<dlan> bjoto: or simply using "b4 shazam", it will also grab the deps and
       apply (as the info is already in cover letter, via changeid..)   [08:48]
<bjoto> dlan: Applying is easy -- the question is if it's useful to the
	maintainers to do so? (Counter argument is that we should try to be as
	aggressively as we can to apply patches/dependencies, to give the
	submitters input.)					        [08:57]
<palmer> bjoto: if the option is "apply dependencies and test a patch set" vs
	 "skip/fail because there's dependencies", then I'd find applying them
	 more useful -- otherwise there's just going to be noise in the test
	 results, and as long as b4 is noticing the dependency then I will too
								        [17:11]

@palmer-dabbelt prefers dlan's approach, so let's implement that!

TODO Use b4-shazam to apply the series, which will catch dependencies as well.