Fix fgraph_from_model with multivariate transformed variables

[ferrine]

def test_do_transformed():
    """There was a bug when the shape check compared value vars
    with actual rv type which can have different shapes"""
    with pm.Model() as m:
        pm.ConstantData("data_std", 1)
        # value var has different number of dimensions
        pm.LKJCorr("c", 4, 1.0)
        # value var has different shape
        pm.Dirichlet("d", [1, 2])
    m2 = do(m, {"data_std": 2})
    assert m2["data_std"].eval() == 2

---

📚 Documentation preview 📚: https://pymc--6924.org.readthedocs.build/en/6924/

[codecov]

Codecov Report

Merging #6924 (b05c04b) into main (9227827) will increase coverage by 0.01%.
Report is 4 commits behind head on main.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6924      +/-   ##
==========================================
+ Coverage   92.16%   92.17%   +0.01%     
==========================================
  Files         100      100              
  Lines       16839    16839              
==========================================
+ Hits        15519    15522       +3     
+ Misses       1320     1317       -3     

Files Changed	Coverage Δ
pymc/model/fgraph.py	96.90% <100.00%> (ø)

... and 1 file with indirect coverage changes