Don't repeat inner graphs with multiple outputs in debugprint

[ricardoV94]

Description

import pytensor
import pytensor.tensor as pt
x = pt.vector("x")
outs, _ = pytensor.scan(
    fn = lambda x: (
        pt.exp(pt.random.normal(x)),
        pt.log(pt.random.normal(x)), 
        pt.cos(pt.random.normal(x)),
    ),
    sequences=[x],
)
fn = pytensor.function([x], outs)
pytensor.dprint(fn)

for{cpu,scan_fn}.0 [id A] 5
 |Shape_i{0} [id B] 0
 | |x [id C]
 |Subtensor{int64:int64:int8} [id D] 4
 | |x [id C]
 | |ScalarFromTensor [id E] 3
 | | |Elemwise{Composite{Switch(LE(i0, i1), i1, i2)}} [id F] 2
 | |   |Shape_i{0} [id B] 0
 | |   |TensorConstant{0} [id G]
 | |   |TensorConstant{0} [id H]
 | |ScalarFromTensor [id I] 1
 | | |Shape_i{0} [id B] 0
 | |ScalarConstant{1} [id J]
 |Shape_i{0} [id B] 0
 |Shape_i{0} [id B] 0
 |Shape_i{0} [id B] 0
 |RandomGeneratorSharedVariable(<Generator(PCG64) at 0x12E952EA0>) [id K]
 |RandomGeneratorSharedVariable(<Generator(PCG64) at 0x12E89F4C0>) [id L]
 |RandomGeneratorSharedVariable(<Generator(PCG64) at 0x12E8E4740>) [id M]
for{cpu,scan_fn}.1 [id A] 5
for{cpu,scan_fn}.2 [id A] 5

Inner graphs:

for{cpu,scan_fn}.0 [id A]
 >Elemwise{exp,no_inplace} [id N]
 > |normal_rv{0, (0, 0), floatX, False}.1 [id O]
 >   |*1-<RandomGeneratorType> [id P] -> [id K]
 >   |TensorConstant{[]} [id Q]
 >   |TensorConstant{11} [id R]
 >   |*0-<TensorType(float64, ())> [id S] -> [id D]
 >   |TensorConstant{1.0} [id T]
 >Elemwise{log,no_inplace} [id U]
 > |normal_rv{0, (0, 0), floatX, False}.1 [id V]
 >   |*2-<RandomGeneratorType> [id W] -> [id L]
 >   |TensorConstant{[]} [id Q]
 >   |TensorConstant{11} [id R]
 >   |*0-<TensorType(float64, ())> [id S] -> [id D]
 >   |TensorConstant{1.0} [id T]
 >Elemwise{cos,no_inplace} [id X]
 > |normal_rv{0, (0, 0), floatX, False}.1 [id Y]
 >   |*3-<RandomGeneratorType> [id Z] -> [id M]
 >   |TensorConstant{[]} [id Q]
 >   |TensorConstant{11} [id R]
 >   |*0-<TensorType(float64, ())> [id S] -> [id D]
 >   |TensorConstant{1.0} [id T]

for{cpu,scan_fn}.1 [id A]
 >Elemwise{exp,no_inplace} [id N]
 >Elemwise{log,no_inplace} [id U]
 >Elemwise{cos,no_inplace} [id X]

for{cpu,scan_fn}.2 [id A]
 >Elemwise{exp,no_inplace} [id N]
 >Elemwise{log,no_inplace} [id U]
 >Elemwise{cos,no_inplace} [id X]

The last two entries of inner-graphs are completely redundant, since the information is already contained in the first inner graph. This will be even more relevant after #121 as multiple output Composites will also be printed like this.

Does anyone know a good reason for repeating the outputs in inner graphs? Maybe inplace or view info?