cumulus-primitives-parachain-inherent: Split into two crates (#2803)

This splits `cumulus-primitives-parachain-inherent` into two crates, the
previous `cumulus-primitives-parachain-inherent` and a new
`cumulus-client-parachain-inherent`. The idea behind this is to move the
`create_at` logic into the client crate. This removes quite a lot of
unrelated dependencies from the runtime std build and thus, makes the
compilation faster. On my Laptop the compilation is goes down by one
minute for `asset-hub-rococo-runtime`. I also assume that the full build
of the entire workspace probably can be speed-up a little bit, because
more stuff can be compiled in parallel.

---------

Co-authored-by: command-bot <>

prdoc/pr_2803.prdoc

@@ -0,0 +1,19 @@
+title: "cumulus-primitives-parachain-inherent: Split into two crates"
+
+doc:
+  - audience: Node Dev
+    description: |
+      This splits `cumulus-primitives-parachain-inherent` into two crates. The new crate is called
+      `cumulus-client-parachain-inherent`. This is done to improve the compile time for runtimes,
+      as they are not required anymore to pull in half of the node side at compile time.
+
+      To migrate your code you need to change
+      `cumulus_primitives_parachain_inherent::ParachainInherentData::create_at` to
+      `cumulus_client_parachain_inherent::ParachainInherentDataProvider::create_at`.
+      Any other code should be compatible. The mocking code also moved to the new client crate and
+      you may need to adapt your imports accordingly. Generally, replacing the old crate with the new
+      crate fix most compile errors resulting from this pull request.
+
+crates:
+  - name: "cumulus-primitives-parachain-inherent"
+  - name: "cumulus-client-parachain-inherent"