[stree] stree to SDFG fixes from NASA branch

[romanc]

Description

This PR brings back a couple of stree -> SDFG fixes from the NASA working branch #2364, which accumulated a bunch of changes last month. In particular, this PR contains

1. consider scalar memlets e.g. when collecting memlets of conditions. our use-case is a conditional inside a triple-nested loop where the conditional depends on a scalar condition that is calculated outside the loop (see test case).
2. connect all source / sink nodes inside a map with empty memlets to the corresponding entry / exit node of the containing map (e.g. see image below of what was wrong before)
3. proper support for (self-assigning) copy nodes
4. ensure unique names of added loop regions (this is enforced by validation)

Support for collecting scalar memlets bleeds out into sdfg.py and state.py. All changes there (outside the schedule tree world) are fully backwards compatible.

---

problematic stree -> sdfg transformation before fix number 2