fix(autogram): Reset even when exception is raised

[ValerianRey]

Before this, the following would fail:

import torch
from torch.nn import Linear
from torchjd.autogram import Engine

m = Linear(2, 3)
t = torch.randn([16, 2])
engine = Engine([m])

output = m(t)
g = engine.compute_gramian(output)  # ok
g = engine.compute_gramian(output)  # ValueError because target edges have been reset, as expected

output = m(t)
g = engine.compute_gramian(output)  # ValueError again, but not expected this time, because in the previous call we didn't reset properly the state and thus everything is now fucked.

Now, we get:

import torch
from torch.nn import Linear
from torchjd.autogram import Engine

m = Linear(2, 3)
t = torch.randn([16, 2])
engine = Engine([m])

output = m(t)
g = engine.compute_gramian(output)  # ok
g = engine.compute_gramian(output)  # ValueError because target edges have been reset, as expected. Still reset state.

output = m(t)
g = engine.compute_gramian(output)  # ok

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/torchjd/autogram/_engine.py	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.