Remove global entry_guards

[juniorz]

For compatibility reasons, we are currently populating this global by calling add_an_entry_guard when a guard is chosen to be used (in an attempt to simulate pick_entry_guards's behaviour).

We should revisit every function that uses entry_guards (and get_entry_guards()) to determine if they still make sense in the context of prop259.

[juniorz]

We should target to not using the global entry_guards (as per today's meeting).

[iapazmino]

Quick search for:
get_entry_guards() shows its only use is when choosing_good_entry_server() to build a circuit [src/or/circuitbuild.c line 2180].
entry_guards (smartlist_t) shows is heavily used inside entrynodes.c to manipulate the list and in a few places outside to build and use a circuit.

[juniorz]

This is a list of functions using the global:

- get_entry_guards
- num_live_entry_guards
- entry_guard_get_by_id_digest
- log_entry_guards
- add_an_entry_guard
- pick_entry_guards
- remove_obsolete_entry_guards
- remove_dead_entry_guards
- remove_all_entry_guards
- entry_guards_compute_status
- entry_guard_register_connect_status
- update_node_guard_status
- entry_guards_set_from_config
- choose_random_entry_impl
- entry_guards_parse_state
- entry_guards_update_state
- getinfo_helper_entry_guards
- entries_retry_helper
- any_bridge_supports_microdescriptors
- entry_guards_free_all

[juniorz]

We need to investigate if these are necessary:

- remove_obsolete_entry_guards
- remove_dead_entry_guards
- entry_guards_compute_status

And how can we replace these:

- getinfo_helper_entry_guards
- entries_retry_helper
- any_bridge_supports_microdescriptors

[juniorz]

We have already done:

- entry_guards_parse_state
- entry_guards_update_state

We could stop adding things to the global, because we already can load and save the USED_GUARDS to/from the state file.

[juniorz]

This is already done:

- entry_guard_get_by_id_digest