feat: impl: Determine if 2 strings of some maximum length are lexicographically consecutive.

[marete]

This will be used in a subsequent PR to correctly determine if 2 Bigtable row keys are adjacent, given that a Bigtable row key must be at most 4KiB in length.

The context is that the function we currently use (which is in the Bigtable client library) to do the check that 2 row keys are adjacent order-wise always assumes that the next string in lexicographic order is obtained by adding a '\0' char, which assumes that row keys may be of unlimited length when in fact they are limited to 4KiB, in Bigtable.

---

This change is 

[prawilny]

I'll review the test after the fixes land.

Reviewed 2 of 3 files at r1, all commit messages.
Reviewable status: 2 of 3 files reviewed, 2 unresolved discussions (waiting on @dopiera)

---

google/cloud/bigtable/emulator/range_set.cc line 616 at r1 (raw file):

  // parts. However, row keys that are at maximum length are the rare
  // exception, presumably.
  successor_to_a = a;

nit: why not copy it in the initialization (with std::string successor_to_a = a)?

---

google/cloud/bigtable/emulator/range_set.cc line 618 at r1 (raw file):

  successor_to_a = a;

  successor_to_a[index_to_inc.value()] += 1;

I think it's wrong.

In the examples, let's assume that char is signed (and 127 is the max value).

According to this function, ConsecutiveStringsOfMaxLen([0, 127], [1,127], 2) would be true, whereas it should be false.

It should be true for ConsecutiveStringsOfMaxLen([0, 127], [1,-128], 2) instead.

I think the cleanest way to finding the successor would be to use find_if() and fill() from https://en.cppreference.com/w/cpp/algorithm.html to find the non-max value, fill everything following it with MIN_CHAR and then increment the value as you do here.
We'd avoid pointer math (std::distance) that way.