I don't think it is absolutely mandatory to run a CFA first unless these scales have not been previously validated/examined for their psychometric properties, and you feel the need to examine their factor structure, reliability, validity etc. first.
If the measure has not been validated in your country with the sample similar to yours, it is recommended to perform a CFA. If your cfa qresults align with the prior cfa results of the measure, in your country you don't need to report your CFA results. Otherwise, you can bring your cfa results as a note to the measure.
Let me know if there is any need for further explanation.
I suppose CFA and LPA are asking different questions, so one is not necessarily dependent on the other. But maybe you are in the position where you have an existing scale, and you want to do a LPA. So you are asking 'are there different groups of people who have different response profiles to the items in the scale?", and this is a reasonable question to ask. But you may also be asking yourself "If I do a LPA will someone not ask me why I didn't do a CFA first?" - and the answer is probably 'Yes'. If you were doing a LPA on a scale that had, say, 3 sub-scales I'd want to see that this was a plausible structure for your data. So, you do a CFA and then a LPA, and someone may reasonably ask "Why have you ignored the latent structure of your items and just chucked them all into a LPA?". I'm not sure how you, or I, would answer that question.
So maybe what you want to do, is rather than doing a CFA and then a LPA, is to do a CFA and a LPA at the same time - that is, combine the two models into what is called a factor mixture model (FMM). The 'factor' bit is the CFA, and the 'mixture' bit is the LPA - it's a really neat idea. So you could do a CFA and then allow the intercepts (item means) or factor means to vary across latent classes.
Here are some good papers
Lubke, G. H., & Muthén, B. (2005). Investigating population heterogeneity with factor mixture models. Psychological Methods, 10(1), 21–39. https://doi.org/10.1037/1082-989X.10.1.21
Gitta Lubke & Michael Neale (2008) Distinguishing Between Latent Classes and Continuous Factors with Categorical Outcomes: Class Invariance of Parameters of Factor Mixture Models, Multivariate Behavioral Research, 43:4, 592-620, DOI: 10.1080/00273170802490673
In terms of how these models can be specified (modelling items or factor), this is an astonishingly good paper
Shaunna L. Clark , Bengt Muthén , Jaakko Kaprio , Brian M. D'Onofrio , Richard Viken & Richard J. Rose (2013) Models and Strategies for Factor Mixture Analysis: An Example Concerning the Structure Underlying Psychological Disorders, Structural Equation Modeling: A Multidisciplinary Journal, 20:4, 681-703, DOI: 10.1080/10705511.2013.824786
and there are some nice demonstrations of how these models can be applied-
Keller, A. C., Igic, I., Meier, L. L., Semmer, N. K., Schaubroeck, J. M., Brunner, B., & Elfering, A. (2017). Testing job typologies and identifying at-risk subpopulations using factor mixture models. Journal of Occupational Health Psychology, 22(4), 503.
Conway, C., Hammen, C., & Brennan, P. A. (2012). A comparison of latent class, latent trait, and factor mixture models of DSM-IV borderline personality disorder criteria in a community setting: Implications for DSM-5. Journal of Personality Disorders, 26(5), 793.
Redican, E., Cloitre, M., Hyland, P., McBride, O., Karatzias, T., Murphy, J., & Shevlin, M. (2022). The latent structure of ICD-11 posttraumatic stress disorder (PTSD) and complex PTSD in a general population sample from USA: A factor mixture modelling approach. Journal of Anxiety Disorders, 85, 102497.
I think it really depends on the domain you based, because different domains have different traditions. If you are going to submit your manuscript to a journal of educational psychology, then CFA is almost a must whenever you use questionnaires. If you aim at an applied linguistics journal, CFA may not be so keenly required by reviewers. But my opinion is to conduct a CFA before any further analysis (including LPA), because this is not so cumbersome, but can add to the robustness of your instrument. So why not do that. ( :