Fix security issue #364 and non-keyword subgraph parsing

[sehe]

(Note: this PR builds in #375 and #374)

Three commits in this PR:

1. ad11d2e8 Fix non-keyword subgraph parsing
2. 82f623f7 test_subgraphs verifies (keyword) subgraphs parse
3. 95b80a93 max_subgraph_nesting_level in read_graphviz_new

The last one actually fixes #364

[TheZ3ro]

Can confirm this fixes #364 correctly

[sehe]

Updated for review comments https://github.com/boostorg/graph/issues/364#issuecomment-2100506381

[sehe]

Added test_subgraph_nesting_limit to pin-down the new behavior.

[jeremy-murphy]

Do you need to rebase this? It has all the changes from the second PR. (Never mind the Drone failures, they're just a network glitch.)

[sehe]

Do you need to rebase this? It has all the changes from the second PR. (Never mind the Drone failures, they're just a network glitch.)

Not necessarily. It's how dependent PRs work in Github (note that I started all PRs with the warning and also explicitly stated it before creating them). Therefore, the second PR already had the same dependency, and it worked "fine":

Rebasing first might simplify the revision graph every so slightly (it's why I prefer linear-only history with ff-only merges). As a late thought: should I have edited my name into the authors for test/graphviz_test.cpp?

[jeremy-murphy]

OK, I'm still a bit confused as to why it is showing all the changes from the second PR, which is merged, but I'll merge this and see what happens.