Jacek Sieka 6080fef1b2 Speed up http client (#633) ...

The http client currently uses an inefficient byte-by-byte copy in its
`BoundedStreamReader` - the real solution here is to get rid of the
copy completely but until then, we can use bulk-copy the data at least.

Ditto `read`, `readN` and similar helpers - this ~doubles throughput
for bulk reading.

Pre:
```
| Small/small    |    0.070s |   1000 | 14226.009 |    0.031 MB |    0.016 MB |    0.434 MB/s |    0.231 MB/s |
| Medium/small   |    1.834s |   1000 |  545.267 | 1000.000 MB |    0.021 MB |  545.267 MB/s |    0.011 MB/s |
| Small/Medium   |    2.238s |   1000 |  446.735 |    0.031 MB | 1000.000 MB |    0.014 MB/s |  446.735 MB/s |
| Medium/Medium  |    2.583s |   1000 |  387.196 | 1000.000 MB | 1000.000 MB |  387.196 MB/s |  387.196 MB/s |
```

Post:

```
| Small/small    |    0.066s |   1000 | 15038.890 |    0.031 MB |    0.016 MB |    0.459 MB/s |    0.244 MB/s |
| Medium/small   |    0.954s |   1000 | 1048.475 | 1000.000 MB |    0.021 MB | 1048.475 MB/s |    0.022 MB/s |
| Small/Medium   |    1.264s |   1000 |  791.318 |    0.031 MB | 1000.000 MB |    0.024 MB/s |  791.318 MB/s |
| Medium/Medium  |    1.615s |   1000 |  619.083 | 1000.000 MB | 1000.000 MB |  619.083 MB/s |  619.083 MB/s |
```

2026-04-07 19:37:35 +02:00

..

bench_bulk_tcp.nim

stream: Bulk reads (fixes #571) (#632)

2026-04-07 15:50:05 +02:00

bench_http_fetch.nim

Speed up http client (#633)

2026-04-07 19:37:35 +02:00