Fix return optimization for nested tuple

[charles-cooper]

The optimization for tuples of base types expected anything inside a
tuple to be an int, but that's not always the case.

Luckily logic introduced in 0605abb can actually be generalized to any
tuple type which has no dynamic data in the ABI encoding, so we just
remove the constraints on the child types.

What I did

Fixes #2390

How I did it

How to verify it

Compile the example in #2390.

Description for the changelog

Fix type error when returning nested tuples.

Cute Animal Picture

[codecov-commenter]

Codecov Report

Merging #2392 (cae5490) into master (db2606d) will decrease coverage by 0.96%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #2392      +/-   ##
==========================================
- Coverage   85.84%   84.88%   -0.97%     
==========================================
  Files          91       91              
  Lines        9022     9022              
  Branches     2152     2152              
==========================================
- Hits         7745     7658      -87     
- Misses        785      866      +81     
- Partials      492      498       +6     

Impacted Files	Coverage Δ
vyper/old_codegen/return_.py	100.00% <100.00%> (ø)
vyper/builtin_functions/functions.py	83.12% <0.00%> (-5.63%)	⬇️
vyper/semantics/types/value/numeric.py	80.43% <0.00%> (-5.44%)	⬇️
vyper/old_codegen/expr.py	75.98% <0.00%> (-4.51%)	⬇️
vyper/semantics/types/function.py	86.02% <0.00%> (-2.21%)	⬇️
vyper/old_codegen/global_context.py	66.92% <0.00%> (-1.73%)	⬇️
vyper/ast/nodes.py	92.12% <0.00%> (-1.47%)	⬇️
vyper/semantics/validation/data_positions.py	90.00% <0.00%> (+3.33%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update db2606d...cae5490. Read the comment docs.

[fubuloubu]

Please add a test for returning a tuple of differnet structs

[charles-cooper]

Please add a test for returning a tuple of differnet structs

@fubuloubu see cae5490 and #2394 (latter explains why I didn't just include the original test cases as is)