tools: add duplicate-samples-generator

[dmatth1]

Adds tool to remote-write requests containing duplicate-timestamp samples and assert how Mimir accounts for them.

What it asserts

3 things per flow:

1. The expected reason on cortex_discarded_samples_total gains exactly one count per duplicate.
2. No other reason moved.
3. cortex_ingester_ingested_samples_total grows by the non-duplicate samples only.

Validation

Run against a local Mimir, both before and after the ingester fixes in #16347 and grafana/mimir-prometheus#1259.

Before the fix - 14 of 24 flows fail (the ones expected):

FAIL: 14 of 24 flows did not match expectations

same-request/same-value/float            same request / in order / same value
    sent                         10 requests / 1000 samples (500 duplicates)
    same-value-for-timestamp     want      500  observed        0  MISMATCH
    ingested samples             want      500  observed     1000  MISMATCH

Nothing recorded as discarded and 1000 counted as ingested where only 500 were stored.

After the fix — all 24 pass:

24 flows across float and histogram
   8 to the distributor at http://localhost:8080
  16 to the ingester at localhost:9095 over gRPC

PASS: all 24 flows matched expectations

The same flow now reads want 500 / observed 500 on both assertions.

Also run against a dev cell through the gateway at replication factor 3, scraping the distributor and all three ingesters: all 24 flows passed, and the stored values were confirmed to be the first-seen ones.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 4e3971f. Configure here.}