Based on my memory of the discharge curve for an 18650, at 3.5V the battery is about 80% discharged, and voltage drops rapidly after that. It's possible the buck/boost converter would give you a shorter battery life than an LDO. I think that's particularly likely if your ESP32 spends any significant time sleeping (switching regulators are typically less efficient than LDOs at very low currents).
But you may find that you need an LDO that will provide more current than the MCP1700. Nominally, 250ma should be plenty, but wifi devices can generate large current spikes when they transmit, and more headroom in the LDO helps deal with those. Also, there's a Youtube video by Andreas Spiess on powering the ESP8266, and he ends up recommending a large (1000µF) electrolytic cap on the 3.3V line to prevent voltage drops caused by the spikes.
Hopefully those with lots of experience with the ESP32 can give you some good advice on this.