Switch from serde to miniserde for Compile Times

[CosmicHorrorDev]

I was wondering how open you would be about switching from serde to miniserde for json deserialization. The overall code change would be tiny for a significant change in compile times. I went ahead and timed a single job compile for each case on my laptop (Intel i5-5200U @ 2.7GHz) and the difference seems pretty significant. (Each run should have been cold and I compiled the version with miniserde before the current master with a full cargo clean between so there shouldn't be any caching affecting this)

serde (Total: 670.52 seconds)

ytop git/master
❯ cargo bloat -j 1 --time --release
  Time Crate
72.11s syn
54.14s serde_derive
46.00s clap
45.14s snafu_derive
43.28s uom
39.36s structopt_derive
28.30s ytop
27.77s derive_more
22.81s regex_syntax
21.64s psutil
20.14s syn_mid
16.12s nix
14.05s regex
12.81s serde
12.42s cc
10.44s tui
10.30s num_bigint
 9.68s serde_json
--- 8< --- snip --- 8< ---
 0.06s cfg_if

miniserde (Total: 466.63)

ytop git/master*
❯ cargo bloat -j 1 --time --release
  Time Crate
53.66s syn
36.12s uom
34.84s derive_more
30.47s structopt_derive
30.23s clap
29.76s snafu_derive
25.32s mini_internal
20.21s ytop
16.07s regex_syntax
13.74s psutil
13.27s syn_mid
11.51s regex
10.64s nix
 7.84s cc
 7.24s proc_macro_error_attr
 6.83s tui
 6.54s num_bigint
 6.20s proc_macro2
 5.28s cassowary
 5.25s crossterm
 5.10s libc
 4.40s chrono
 4.11s fern
 3.79s mio
 3.43s console
 3.20s miniserde
--- 8< --- snip --- 8< ---
 0.06s cfg_if

Even though this was a single job build there should still be a significant difference in multithreaded build times. I just a single job since it seemed like the cleanest metric to choose.