IR: Automatic sanitisation of tuples in IR constructors

[mlange05]

Just`` testing for now....

Note, this is based on #349, as it requires pydantic version >=2.0.

Following the Pydantic upadte, this PR now smoothes over the type-checking of the base IR nodes, in particular InternalNode, and adds "smart" conversion of .body tuples to the desired internal format. In particular, this now means that erroneous tuples (eg. nested, or containing None instances) are now automatically fixed, before being passed to the pydantic validators. This in turn will allow a more liberal use of Transformers for in-traversal replacements (the real point of this change!).

In more detail:

• Added a small _sanitise_tuple utility method to loki.ir.nodes
• Explicitly typed .body as Tuple[Union[Node, Scope], ...] to allow nested Subroutine objects (eg. internal procedures)
• Added auto-casting tests to IR tests for Loop and Conditional, and added test for Section, which tests only the inherited behaviour from InternalNode

[github-actions]

Documentation for this branch can be viewed at https://sites.ecmwf.int/docs/loki/350/index.html

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.35%. Comparing base (536908b) to head (561d748).
Report is 12 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #350      +/-   ##
==========================================
+ Coverage   95.33%   95.35%   +0.01%     
==========================================
  Files         171      171              
  Lines       36543    36801     +258     
==========================================
+ Hits        34837    35090     +253     
- Misses       1706     1711       +5     

Flag	Coverage Δ
lint_rules	96.39% <ø> (ø)
loki	95.33% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[reuterbal]

This looks fantastic, many thanks, and we finally gain the flexibility of being more forgiving in the tuple nesting in the constructor while ensuring correct IR nodes.

One question about a potential change in default behaviour for Conditional and whether this is intentional, otherwise good to go!

[reuterbal]

Is that a change in behaviour for the default value of else_body?
I think so far we used else_body=None as indicator that there isn't an else branch in the conditional - as opposed to an empty else branch (which has the same result but different on a string level).

As in, will ir.Conditional(condition=<...>, body=<...>) now be represented as the following?

IF (<...>) THEN
  <...>
ELSE
ENDIF

[mlange05]

Nope, we're doing a plain old if o.else_body: which evaluates to the same thing:
https://github.com/ecmwf-ifs/loki/blob/main/loki/backend/fgen.py#L633

[mlange05]

@reuterbal I just noted that we now have a hard dependency on pydantic>=2.0 with this, so please let me quickly update the setup file....

[mlange05]

Ok, fix pushed. Please note that existing local checkouts might fail after this - to remedy, a simple pip install -e . in the loki directory should suffice to get pydantic updated.