`too-deep-object` lint tool reports multiple defects for single nesting chain and emits incorrect depth number

Problem

too-deep-object reports several defects along a single nesting chain (a cascade) and prints a depth number that does not match the rule's threshold, confusing the user.

Root cause

src/main/resources/org/eolang/lints/design/too-deep-object.xsl:13 (selector) and :28 (message):

• The selector excludes certain wrapper objects (Φ.tuple, Φ.bytes, Φ.number, Φ.string) from the depth count, while the message prints the raw count(ancestor::o) — so the number in the text (e.g. "17 levels") differs from the filtered depth that triggered the rule.
• Because of the counting difference, more than one node per chain can match the not(o[count(ancestor::o[...]) > 12]) guard, producing a cascade of defects for one chain (observed: defects at depths 13/14/15/17 for a 15-level chain).
• For data objects the message prints eo:escape(@base) which yields an empty/∅ name.

Minimal example

+package x
+architect yegor@256.com
+version 0.0.1
+home https://github.com/objectionary
+spdx SPDX-FileCopyrightText: Copyright (c) 2026 Objectionary.com
+spdx SPDX-License-Identifier: MIT

[] > main
  [] > a
    [] > b
      [] > c
        [] > d
          [] > e
            [] > f
              [] > g
                [] > h
                  [] > i
                    [] > j
                      [] > k
                        [] > l
                          [] > m
                            1 > @

Expected behavior

• The depth number in the message must use the same (filtered) metric as the selector.
• Only the deepest offending node of a branch should be reported.
• Use @name (or a clear label) instead of @base for formations.

Related

• #155 (the lint itself, merged); the cascade/off-by-N behavior is a new regression on top of it.
• design/ category siblings (broad-scope, excessive-visibility).

Start with src/main/resources/org/eolang/lints/design/too-deep-object.xsl, especially the selector at line 13 and message at line 28. Align the reported depth with the filtered metric, report only the deepest offending node in a branch, and use @name or another clear label for formations; verify the minimal example no longer produces cascaded defects.