fix(posting): make rollup failures diagnosable in logs

[matthewmcneely]

Description

Two problems make a failing rollup nearly impossible to triage from logs. Both
turned up while looking at #9794, where the reporter's log line was truncated
right before the useful part.

1. Reversed format arguments in incrRollupi.Process

glog.Warningf("Error rolling up key [%v]: %v", err, key)

err and key are swapped, so the message reads Error rolling up key [<the error>]: <the key>. The key also rendered as a decimal byte slice, which can't
be pasted into dgraph debug --lookup. Now %x with the arguments in the right
order.

2. The pIterator.seek error message dumps the whole posting list

List.iterate formatted l.Print() into its error, and Print() dumps the
entire posting list plus mutation map with posting values included. On a
multi-part list holding large values that's a multi-megabyte log line. Since
ReadPostingList re-queues a key for rollup on every read while deltas remain,
and doRollup retries a given key roughly every 10 seconds, the same giant line
repeats for as long as the key stays broken.

Worse for triage: errors.Wrapf appends the cause after the format string, so
the actual error is the first thing to get cut off. seek has only two failure
modes and they have completely different root causes, so losing that suffix
loses the diagnosis:

deleteBelowTs (%d) must be greater than the minTs in the list (%d)
cannot read initial list part for list with base key %s

This PR adds List.debugString(), which reports the key (hex), minTs,
maxTs, splits, and mutation-layer counts, with no posting values, and uses it
on the three seek error paths in iterate and findPostingWithItr. The two
paths in findPostingWithItr were formatting mutationMap.print(), which has
the same unbounded-value problem.

Print() is unchanged and stays available for interactive debugging.

Sample of the new output shape:

cannot initialize iterator when calling List.iterate key: 000764617461..., minTs: 58256931,
maxTs: 58550095, splits: [1 8523883114575535643], mutationMap: {committed: 1, proposed: false,
deleteAllMarker: 18446744073709551615}: could not read list part with key ...: Key not found

Logging only. No behavior change.

Refs #9794.

Checklist

• The PR title follows the
  Conventional Commits syntax, leading
  with fix:, feat:, chore:, ci:, etc.
• Code compiles correctly and linting (via trunk) passes locally

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}