[iddqd-benches] expand benchmark suite

[sunshowers]

Add benchmarks for bulk_insert, churn, iter, shrink_to_fit, and a
ref_mut microbench alongside the existing get sweep.

These benches compile and run against the current HashMap-backed
ItemSet, establishing a baseline before any storage-layout changes.
Criterion medians on a Ryzen 7950x:

get (point lookup, ns):
    size  std_hash  std_btree  id_hash_own  id_hash_brw  id_ord_own  id_ord_brw
       1      1.75       2.02         6.00         5.97        6.90        6.89
      10      2.81       1.81        10.43        10.38        7.77        7.83
     100      3.14       9.07        14.09        13.82       25.88       25.72
      1k      4.01      13.84        16.91        16.97       41.52       39.40
     10k      6.65      24.79        23.12        21.74       75.48       66.99
     50k     22.91      43.45        38.61        37.95       90.01      104.21
    100k     28.91      68.69        51.67        53.79      136.18      106.45
    500k    126.70     299.95       328.58       320.65      1248.5      1209.2
      1M    141.33     873.01       346.78       366.89      1230.0      1223.1

bulk_insert (fill N records):
  size     std_hash    std_btree      id_hash       id_ord
   100      1.59 µs      1.75 µs      2.36 µs      9.69 µs
   10k    163.05 µs    299.17 µs    262.71 µs      2.17 ms
  100k      1.55 ms      3.15 ms      2.81 ms     26.78 ms

churn (1000 remove+reinsert pairs, µs):
  size   std_hash  std_btree   id_hash    id_ord
   100      14.72      40.97     30.73    139.45
   10k      19.14      43.68     50.81    282.23
  100k      24.20      53.13     63.75    304.27

iter (full pass):
  size     std_hash    std_btree      id_hash       id_ord
   100     46.11 ns     82.96 ns     45.79 ns    289.07 ns
   10k      5.69 µs      9.28 µs      5.75 µs     39.01 µs
  100k     62.40 µs    136.00 µs     53.96 µs    480.80 µs

shrink_to_fit (fill, scatter ~50% holes, compact):
  size     std_hash      id_hash       id_ord
   100    295.54 ns    643.52 ns    273.63 ns
   10k     26.35 µs    348.79 µs     24.38 µs
  100k    283.52 µs     36.03 ms    251.36 µs

ref_mut/id_ord_map: 13.76 ns