Partial "answer":
In the beginning there were the double (32-bit?) and the char, and everything was peaceful, then they ate of the forbidden fruit and saw that there were special use-cases where new types would be beneficial, large images (gray or colour) could be much easier handled in a smaller bit-format, so they started to multiply, the int8, then the int16 and all their signed and unsingned siblings, and suddendly you could no longer index arrays with flints - but you got to use multidimensional arrays, then there was a struct and cell and they were ever so handy, then there was a function-handle that came walking by and made space...
(others could join in...)
Every data-type have its useful use-cases where they are ever so handy, when working with large images that have 8-bit intensities it can be handy not having to convert them to 32(64?!)-bit double data, just to not waste space. Structs and cell-arrays are very convenient types to bundle data of various types that should "go together" (for one reason or other). Time is a notoriously messy data-type to handle well and generally (daylight-savings-time, midnight and elapsed times and time-zones and on and on...) so a proper set of functions for handling all of that.
