perf: only compute start offset for overlong lines (#5811)
Moves the computation of the `start_offset` for overlong lines to just
before the result is returned. There is a slight overhead for overlong
lines (double the work for the first `limit` characters).
In practice this results in a speedup on the CPython codebase. Most
lines are not overlong, or are not enforced because the line ends with a
URL, or does not contain whitespace. Nonetheless, the 0.3% of overlong
lines are a lot compared to other violations.
### Before
_Selected W505 and E501_
_All rules_
### After
_Selected W505 and E501_
_All rules_
CPython line statistics:
- Number of Python lines: 867.696
- Number of overlong lines: 2.963 (0.3%)
<details>
Benchmark selected:
```shell
cargo build --release && hyperfine --warmup 10 --min-runs 50 \
"./target/release/ruff ./crates/ruff/resources/test/cpython/ --no-cache -e --select W505,E501"
```
Benchmark all:
```shell
cargo build --release && hyperfine --warmup 10 --min-runs 50 \
"./target/release/ruff ./crates/ruff/resources/test/cpython/ --no-cache -e --select ALL"
```
Overlong lines in CPython
```shell
cargo run -p ruff_cli -- check crates/ruff/resources/test/cpython/Lib --no-cache --select=E501,W505 --statistics
```
Total Python lines:
```shell
find crates/ruff/resources/test/cpython/ -name '*.py' | xargs wc -l
```
</details>
(Performance tested on Mac M1)
