Draw the Moon Phase

JavaScript (ES6), 121 ... 103 92 bytes

f=(n,i=128)=>i--?f(n,i)+(i%16?'':`
`)+'.#.'[i%16+4*n-~-'31000013'[n&1?i>>4:1]*~-(n&2)>>4]:''

Demo

f=(n,i=128)=>i--?f(n,i)+(i%16?'':`
`)+'.#.'[i%16+4*n-~-'31000013'[n&1?i>>4:1]*~-(n&2)>>4]:''

o.innerHTML=f(0)

<input type="range" oninput="o.innerHTML=f(this.value)" value=0 min=0 max=7><pre id=o>

Jelly,  43 32 bytes

-7 bytes moving from bit-mask to comparison-mask
-2 bytes with some tacit programming improvements
-1 byte moving to the southern hemisphere
-1 byte - use the all-new ɓ chain separator ...its first ever use!

“>,##‘m0Dm€0ż@€4Ḷ¤Œṙɓ%4>C¹⁸:4¤?Y

Uses the characters 0 for . and 1 for #.

All cases in a test suite at Try it online!

How?

Builds a mask capturing the first four phases, and rather than reversing the lines complements the resulting values of the modulo-4 phase result when the phase div-4 is non-zero.

I originally built a bit mask, but the mask values were 0,8,12, and 14 - 0000, 1000, 1100, and 1110 - these have phase leading ones - so a comparison-mask could be used instead.

“>,##‘m0Dm€0ż@€4Ḷ¤Œṙɓ%4>C¹⁸:4¤?Y - Main link 1: number phase
“>,##‘                           - code-page index literal [62,44,35,35]
      m0                         - reflect -> [62,44,35,35,35,35,44,62]
        D                        - decimalise -> [[6,2],[4,4],[3,5],[3,5],[3,5],[3,5],[4,4],[6,2]]
         m€0                     - reflect €ach -> [[6,2,2,6],[4,4,4,4],[3,5,5,3],[3,5,5,3],[3,5,5,3],[3,5,5,3],[4,4,4,4],[6,2,2,6]]
                 ¤               - nilad and link(s) as a nilad:
               4                 -   literal 4
                Ḷ                -   lowered range -> [0,1,2,3]
            ż@€                  - zip (reverse @rguments) for €ach -> [[[0,6],[1,2],[2,2],[3,6]],[[0,4],[1,4],[2,4],[3,4]],[[0,3],[1,5],[2,5],[3,3]],[[0,3],[1,5],[2,5],[3,3]],[[0,3],[1,5],[2,5],[3,3]],[[0,3],[1,5],[2,5],[3,3]],[[0,4],[1,4],[2,4],[3,4]],[[0,6],[1,2],[2,2],[3,6]]]
                  Œṙ             - run-length decode -> [[0,0,0,0,0,0,1,1,2,2,3,3,3,3,3,3],[0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3],[0,0,0,1,1,1,1,1,2,2,2,2,2,3,3,3],[0,0,0,1,1,1,1,1,2,2,2,2,2,3,3,3],[0,0,0,1,1,1,1,1,2,2,2,2,2,3,3,3],[0,0,0,1,1,1,1,1,2,2,2,2,2,3,3,3],[0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3],[0,0,0,0,0,0,1,1,2,2,3,3,3,3,3,3]]
                                    -   i.e.: 0000001122333333  -
                                              0000111122223333  - Marking out the separate
                                              0001111122222333  - regions as filled up by
                                              0001111122222333  - the phases of the moon in
                                              0001111122222333  - the southern hemisphere.
                                              0001111122222333  -
                                              0000111122223333  -
                                              0000001122333333  -
                    ɓ            - dyadic chain separation & swap arguments, call that m
                     %4          - phase mod 4
                       >         - greater than? (vectorises across m) 1 if so 0 if not
                              ?  - if:
                             ¤   -   nilad followed by link(s) as a nilad:
                          ⁸      -     link's left argument, phase
                           :4    -     integer divide by 4
                        C        - then: complement
                         ¹       - else: identity (do nothing)
                               Y - join with newlines
                                 - implicit print

Haskell, 98 90 bytes

f i=do{a<-[3,1,0,0,0,0,1,3];[".#."!!div(i*4+x+[0,a-1,0,1-a]!!mod i 4)16|x<-[0..15]]++"\n"}

It loops through the rows and columns using the list monad (a do block for rows and a list comprehension for columns) and determines which character for each cell using a expression of the input (i), offset value for the row (a), and column index (x).

Saved 8 bytes by simplifying the subexpression for the true offset.