It's likely that you don't have to simulate even a single cell at high resolution to be able to simulate how an organism would grow. There are numerical shortcuts.
For example today we can already predict the color of the eyes and other phenotype from the DNA.
If you are able to observe enough samples of cell growth and their associated DNA, you probably can model and predict the statistics of a cell from their DNA. Because the cell is itself the result of a lot of chemical processes, the law of large number will help smooth those statistics.
Given that we have a lot of cells, the collective behavior is probably entirely governed by these statistics.
You seriously underestimate the continuous growth of computer power. And quantum computers after, which are perfect for simulating chemical reactions.
What was unthinkable 50 years ago, playing chess better than a human, it's now trivial for a $100 device.
And it's not necessarily required that to simulate the growth of a human you'll need to simulate the entirety of chemical reactions in all 50 trillion cells and all that.
It's possible I underestimate, but I have worked in all the relevant fields of simulation, ~20 years of running various simulations on large HPC, built the largest instance of folding@home using idle cycles inside google data centers, published papers simulating proteins, developed infrastructure to process the voluminous data, etc, etc. Quantum computing remains fantasy (in terms of being useful for science).
It's unlikely even if we improved computing hardware many orders of magnitude beyond all reasonable predictions, that the calculations would be able to simulate all the necessary details; most of our simulations now are based on many approximations due to hardware limitations.
As to the question of "what level of fidelity is required to turn a FASTQ of somebody's genome into an accurate model of the resulting human, with some sort of realistic environment also provided", that's so far beyond what is even remotely comprehensible it's not worth speculating about in terms of science fact; it's just fiction.
For example today we can already predict the color of the eyes and other phenotype from the DNA.
If you are able to observe enough samples of cell growth and their associated DNA, you probably can model and predict the statistics of a cell from their DNA. Because the cell is itself the result of a lot of chemical processes, the law of large number will help smooth those statistics.
Given that we have a lot of cells, the collective behavior is probably entirely governed by these statistics.