The origin of the correlations between mass, morphology, quenched fraction, and formation history in galaxies is difficult to define, primarily due to the uncertainties in galaxy star-formation histories. Star-formation histories are better constrained for higher redshift galaxies, observed closer to their formation and quenching epochs. Here we use non-parametric star-formation histories and a nested sampling method to derive constraints on the formation and quenching timescales of quiescent galaxies at $0.7<z<2.5$. We model deep HST grism spectroscopy and photometry from the CLEAR (CANDELS Lyman$-alpha$ Emission at Reionization) survey. The galaxy formation redshifts, $z_{50}$ (defined as the point where they had formed 50% of their stellar mass) range from $z_{50}sim 2$ (shortly prior to the observed epoch) up to $z_{50} simeq 5-8$. editone{We find that early formation redshifts are correlated with high stellar-mass surface densities, $log Sigma_1 / (M_odot mathrm{kpc}^{-2}) >$10.25, where $Sigma_1$ is the stellar mass within 1~pkpc (proper kpc). Quiescent galaxies with the highest stellar-mass surface density, $logSigma_1 / (M_odot mathrm{kpc}^{-2}) > 10.25$, } show a textit{minimum} formation redshift: all such objects in our sample have $z_{50} > 2.9$. Quiescent galaxies with lower surface density, $log Sigma_1 / (M_odot mathrm{kpc}^{-2}) = 9.5 - 10.25$, show a range of formation epochs ($z_{50} simeq 1.5 - 8$), implying these galaxies experienced a range of formation and assembly histories. We argue that the surface density threshold $logSigma_1/(M_odot mathrm{kpc}^{-2})>10.25$ uniquely identifies galaxies that formed in the first few Gyr after the Big Bang, and we discuss the implications this has for galaxy formation models.