I think we are kind of good, the performance is halved but with the proxy inside we need to do twice as much work.
The results below come from casual runs on my laptop.

Baseline

$ make bench
wrk http://localhost:8080 -d 10 -t 1 -c 100
Running 10s test @ http://localhost:8080
  1 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   541.77us  205.59us   3.62ms   82.15%
    Req/Sec   124.84k     5.30k  139.03k    73.00%
  1242149 requests in 10.00s, 151.63MB read
Requests/sec: 124198.21
Transfer/sec:     15.16MB

Envoy

$ make bench-proxy
wrk http://localhost:3128 -d 10 -t 1 -c 100
Running 10s test @ http://localhost:3128
  1 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     1.36ms  766.59us  14.56ms   81.36%
    Req/Sec    69.22k     2.08k   72.82k    88.12%
  695662 requests in 10.10s, 108.14MB read
Requests/sec:  68868.27
Transfer/sec:     10.71MB

Forwarder

$ make bench-proxy
wrk http://localhost:3128 -d 10 -t 1 -c 100
Running 10s test @ http://localhost:3128
  1 threads and 100 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     1.55ms  417.42us  15.75ms   88.13%
    Req/Sec    61.44k     2.04k   62.97k    98.00%
  611214 requests in 10.00s, 74.61MB read
Requests/sec:  61106.71
Transfer/sec:      7.46MB

Fixes #676