tuning socket i/o buffer size upto 16384 to get more throughputs

[inter169]

in my user cases there are numbers of requests (around 200) pipelined, and the total size of such responses is much more than 8192 bytes, so in the lower tier it called a bit more times of the syscal 'recvfrom', and the outputs from strace is below:
12207 17:02:03.556594 sendto(289, "*5\r\n$9\r\nZREVRANGE\r\n$21\r\nfeed:user:*****:posts\r\n$1\r\n0\r\n$4\r\n1001\r\n$10\r\nwithscores\r\n", 81, 0, NULL, 0 <unfinished ...>
12207 17:02:03.560635 recvfrom(289, <unfinished ...>
12207 17:02:03.561209 <... recvfrom resumed> "*2000\r\n$24\r\n ..., 8192, MSG_DONTWAIT, NULL, NULL) = 8192 <0.000559>
12207 17:02:03.562899 recvfrom(289, <unfinished ...>
12207 17:02:03.563336 <... recvfrom resumed> "a70 ..., 8192, MSG_DONTWAIT, NULL, NULL) = 8192 <0.000432>
12207 17:02:03.565026 recvfrom(289, <unfinished ...>
12207 17:02:03.565464 <... recvfrom resumed> "bf5 ..., 8192, MSG_DONTWAIT, NULL, NULL) = 8192 <0.000433>
12207 17:02:03.567184 recvfrom(289, <unfinished ...>
12207 17:02:03.567765 <... recvfrom resumed> "459 ..., 8192, MSG_DONTWAIT, NULL, NULL) = 8192 <0.000577>
12207 17:02:03.569528 recvfrom(289, <unfinished ...>
12207 17:02:03.570149 <... recvfrom resumed> "32b ..., 8192, MSG_DONTWAIT, NULL, NULL) = 8192 <0.000617>
12207 17:02:03.572850 <... recvfrom resumed> "f89 ..., 8192, MSG_DONTWAIT, NULL, NULL) = 8192 <0.000681>
12207 17:02:03.574861 recvfrom(289, <unfinished ...>
12207 17:02:03.575534 <... recvfrom resumed> "a0- ..., 8192, MSG_DONTWAIT, NULL, NULL) = 8192 <0.000669>
12207 17:02:03.577385 recvfrom(289, <unfinished ...>
12207 17:02:03.577878 <... recvfrom resumed> "-f98 ..., 8192, MSG_DONTWAIT, NULL, NULL) = 1247 <0.000488>

and the hiredis is using the read buffer size hardcoded as 16384
https://github.com/antirez/redis/blob/unstable/deps/hiredis/hiredis.c

this PR tuned the read buffer size upto 16384, also the write buffer size was upto 16384, to take more throuthputs,
please review,

thanks,
harper