Chore: Add benchmark for `KsymIter::from_kallsyms()`

[patnebe]

Context

• Some performance improvements to KsymIter

Changes

• Added the first benchmark
• Reduce the number of heap allocations in KsymIter::from_kallsyms() by
  • replacing lines() with read_line() + a reusable buffer.
  • using iterators directly instead of collecting the result of line.split() into a vector.

Test Plan

• Unit tests
• Benchmark results on a test kallsyms file containing 1013568 lines. I consistently observed similar improvements for a file with 675712 entries. However, it's worth noting that I don't see the same performance gains when I run this against a file with 337856 entries.

  Benchmarking Read /proc/kallsyms/Measure read all ksyms: Warming up for 5.0000 s
  Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 8.4s, or reduce sample count to 50.
  Read /proc/kallsyms/Measure read all ksyms
                      time:   [84.522 ms 84.705 ms 84.871 ms]
                      change: [-24.298% -24.049% -23.840%] (p = 0.00 < 0.05)
                      Performance has improved.
  Found 13 outliers among 100 measurements (13.00%)
  3 (3.00%) low severe
  5 (5.00%) low mild
  4 (4.00%) high mild
  1 (1.00%) high severe