[CFX-7834] refactor(pipeline): validate count flags at parse time

[ajalon1]

RATIONALE

The pipeline list commands did no flag validation at all — the workload family validated, this family didn't, so the same mistake failed differently depending on which command ate it. The latent bugs this closes:

• --limit 0 (or negative) silently returned fewer rows than asked. The old query builder omitted the limit param when it was <= 0, so the server's default page size applied: a script requesting 100 rows got the default, with no error and nothing in the output saying so.
• A negative --offset silently restarted results from the beginning. Same omission: the param was dropped, so every page request returned page 1. A paging loop with a computed offset that dipped negative re-fetched duplicate rows forever instead of failing.
• pipeline run task logs --tail -5 sent tail_lines=-5 to the live-log endpoint. The negative value was forwarded verbatim once the flag was set, producing a server-side failure mid-command — after authentication, deep past the point of feedback.
• No parse-time gate for CI. Any of the above burned a full auth + network round trip before surfacing; now cobra rejects them during flag parsing.

Migrating onto internal/countflags (from part 1, #858) makes the fix mechanical rather than per-command.

CHANGES

• --limit -> countflags.PositiveInt on dr pipeline list (default 50) and version/run/image/input/schedule list (default 100).
• --offset -> countflags.NonNegativeInt (default 0) on all six list commands.
• --tail on dr pipeline run task logs -> NonNegativeInt, keeping 0 as "no limit".
• New invalid-limit/invalid-offset reject tests per list command and an invalid---tail test, in the existing runCmd style.

TESTING

task lint (all GOOS legs) and task test (race + coverage) pass.

NOTES

Part 2/4 of a stacked review. Based on #858 (base = its branch), so this diff shows only the pipeline changes. Invalid values now error at parse time instead of reaching the API. (Result truncation for large limits is a separate fix, part 4 #861.)

---

Note

Low Risk
CLI-only flag validation and tests; behavior is stricter for invalid pagination/tail values with no changes to auth or server contracts beyond not sending bad query params.

Overview
Pipeline list commands and run task logs now validate pagination and tail flags during Cobra parsing via internal/countflags, instead of plain IntVar flags that could reach the API with bad values.

--limit uses PositiveInt (rejects 0 and negatives) on pipeline list, and on image/input/run/schedule/version list commands with the same defaults as before. --offset uses NonNegativeInt on those list commands. pipeline run task logs --tail uses NonNegativeInt so negatives fail before auth/network; --tail 0 remains “no limit.”

Each touched command gains tests that invalid --limit, --offset, or --tail fail at parse time with the expected error messages.

^{Reviewed by Cursor Bugbot for commit 23c9407. Configure here.}

[datarobot-pr-review-router]

🎫 Jira: CFX-7834 — Refactor --limit/--offset flags to custom pflag.Value (move validation to parsetime)

[sunny2get]

LGTM - thanks for covering this.