The bit depth is the number of bits per sample. In audio, that corresponds to the SNR and maximum dynamic range of the music. When reading a digital file, a DAC has to approximate the ~infinite values of the analog sine wave. The fewer bits there are per sample, the more rounding is necessary. This creates "quantization noise", junk that's not in the original signal. Each bit per sample decreases the amount of rounding, and decreases the volume of the quantization noise by about 6 dB. So a 16 bit depth file will generate quantization noise about 96 dB below the maximum volume. This is enough for all recordings in 95% of all situations (only very loud playback of quiet recordings, or extensive use of digital volume control, will make it audible). 24 bit audio gives 144 dB of SNR, which is more than any DAC can manage (the inherent noise added by the electronics is higher than this, usually SNR is ~120 dB or 20 effective bits at most).
When converting from CD to FLAC, you want to convert it at 16 bits. That's what the CD is, and increasing it will just pad the files with extra zeroes and waste space.
However, if your DAC supports it, there's no reason not to set playback to 24 bits, even if your audio is 16 bits. This will automatically add the extra zeroes, requires no wasted space, and allows for a lot of digital volume attenuation without making the quantization noise audible.
Presumably "Auto" will set it to 16 bits for ripping CDs. As in, it sets it to whatever the source file is. So if you were converting 24 bit WAV files, it would set it to 24 bits.
