[FIX] Avoid raised error for varied step_size parameter during Predict_insample()

[JQGoh]

Changes

• Fixes https://github.com/Nixtla/neuralforecast/issues/866 issue
• Add optimizer, optimizer_kwargs params to TimeLLM and NLinear models. The recently introduced models did not consider these parameters because the optimizer related args were also introduced recently.

Rationale for fix

• Instead of filtering the generated cutoffs by first assuming step_size = 1, we directly impute the series provided by the users.
• As compared to the approach considerd in earlier work, we still pass effective step_size parameter to _insample_times() to validate np.divmod(sizes - h, step_size), ensuring modulo operation = 0.

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[JQGoh]

@quest-bot loot #866

[quest-bot[bot]]

Quest PR submitted!

@JQGoh, you are attempting to solve the issue and loot this Quest. Will you be successful?

---

Questions? Check out the docs.