I am not sure I understand what you are asking for, so sorry if I only confuse you.
Analytically you get the marginal of the ith dimension by looking at a Gaussian with mean = mu[i] and sigma = cov[i,i], where mu and cov are the posterior mean and covariance parameters of your 100th dimensional gaussian. If you have samples from the posterior, you should be able to simply select trace["my_gaussian"][:, i] where my_gaussian is the name you gave to the 100th dimensional gaussian in your pymc3 model.
If this does not answer your question, maybe it would be helpful to have a snippet of your model to understand exactly what you are looking for.