feat: measure host-header-mismatch for HTTP endpoint (no TLS)

[DeagleGross]

since I see an enormous difference between Kestrel and Http.sys RPS perf on host-header mismatch (300k vs 5k), I collected network traces and I can see Kestrel does not close the connection compared to HTTP.SYS - therefore having a better perf.

I decided to measure what perf will be on http only (no TLS) and current PR covers this part.

kestrel:

Stat	Value
App Max Process CPU Usage (%)	96
App Max Global CPU Usage (%)	97
Load Max Process CPU Usage (%)	31
Load Max Global CPU Usage (%)	42
Requests/sec	400,344
Requests	6,044,937
Bad responses	6,044,937
Socket errors	0

http.sys (connections 32)

Stat	Value
App Max Process CPU Usage (%)	0
App Max Global CPU Usage (%)	35
Load Max Process CPU Usage (%)	40
Load Max Global CPU Usage (%)	51
Requests/sec	126,891
Requests	1,915,963
Bad responses	1,915,963
Socket errors	0

http.sys (connections 256)

Stat	Value
App Max Process CPU Usage (%)	0
App Max Global CPU Usage (%)	58
Load Max Process CPU Usage (%)	58
Load Max Global CPU Usage (%)	72
Requests/sec	177,557
Requests	2,681,080
Bad responses	2,681,080
Socket errors	0