refactor: walk a subtree in steps, with its termination rules in one place - #46

…arkWalk

All pass on the walk before its restructuring:

- sibling-arc: an answer whose name extends the root's last arc
  (.1.3.6.1.2.1.2.2.1.20.1 when walking .1.3.6.1.2.1.2.2.1.2) is outside
  the subtree and ends the walk.
- user-code/walkfn-sets-max-repetitions: a walkFn that changes
  MaxRepetitions does not change the walk's later GetBulk requests.
- user-code/walkfn-sets-app-opts-c: a walkFn that sets the AppOpts 'c'
  option does not turn off the walk's increasing check.

BenchmarkWalk runs WalkAll and BulkWalkAll, as the Netdata Agent's SNMP
collector does, over a 100-row table column on the in-memory agent.

…e place

walk was one labeled loop whose break and continue carried the decisions,
with one case per error status. It is now:

- walk: the loop. It sends the walk's request, has visit judge the
  answer, and acts on the returned walkStep: ask for the objects after
  the answer's last name, get the root (a leaf), or end.
- walkRoot: the default root and the leading dot.
- walker.request: GetBulk, GetNext or Get of the walk's OID.
- walker.visit: the rules for an answer: an empty answer, the error
  statuses (one range, TooBig to InconsistentName, logged by their
  generated names), exception values, the subtree, the increasing check,
  walkFn.
- walker.outside: an object outside the subtree: the leaf fallback, the
  root itself, the end.
- inSubtree: strings.HasPrefix(name, root+".") without building the
  prefix for every object.

MaxRepetitions and AppOpts are still read once per walk. walkFn stays a
parameter: as a field of walker it escapes with the client and moves
WalkAll's closure and results to the heap. The walk's log lines are
guarded by Logger.enabled(). Two dead conditions are gone: the
NoSuchInstance check of the root branch (the exception check returns
first) and the request-type condition of the AppOpts 'c' check (walk
runs only GetNext and GetBulk). Each pinned walk bug has a known-bug
comment where it happens.

No behavior change: goldens and logger output unchanged.

A change that makes the engine loop (a regression or a mutation probe)
used to grow the scenario's transcript until the test timed out, holding
gigabytes when several such runs had no timeout. The transcript now fails
its test once it passes 5000 lines, about a hundred times the longest
scenario (53 lines), so such a change fails in milliseconds.

All pass on the walk before its restructuring:

- leaf/root-echo: a first answer naming the root itself makes the walk
  take the root for a leaf and get it.
- leaf/get-answer-in-subtree: after the leaf fallback, an answer inside
  the subtree is walked and the walk sends a Get for that name (known
  bug: net-snmp's snmpwalk makes one Get).
- not-increasing/app-opts-c-nil: the AppOpts 'c' key turns the
  increasing check off whatever its value, nil included.
- error-status/19-unknown-empty: an answer with an unnamed error status
  and no varbinds ends the walk as an empty answer does.

From the step's reviews:

- The error-status rule names its range, tooBig to inconsistentName,
  in its condition and its comment instead of relying on the noError
  case before it.
- The leaf-root known bug sits where it happens: an endOfMibView first
  answer, or SNMPv1's noSuchName, ends the walk before the fallback.
- A new known-bug comment where the fallback switches to Get: the walk
  keeps sending Gets after it.
- The default root's comment cites X.690 alone; RFC 2578 section 7.1.3
  sets no minimum number of arcs.
- The fallback cites gosnmp/gosnmp#78, which added it; the subtree
  example names the excess names correctly.

No behavior change: goldens and logger output unchanged.