Add a warning whenever lost/missed events occur

[curche]

Brief description

Whenever olly detects any events that couldn't be read before being overwritten in the runtime ring buffer, we print a warning about the same at the end of the run (and refers to the missed events section in readme)

Similar to how hyperfine reports of statistical discrepancies during runs, but this one's simpler and mostly just informative.

Other details

The implementation for now is just using a ref 0 that increments - the value of which doesn't really matter just that its either zero or positive non-zero.
The other alternatives were a) something with effect handlers (but that felt a bit overkill) or b) having a warning only during gc-stats through an updated config (but it seemed like both trace and gc-stats (& future options) could benefit from showing the warning)

I've kept the Printf in lost_events but maybe that could be reconsidered now that the final result gives a warning for lost events - the benefit being that excessive no of lost events will not affect the target executable's output (if any)

Sample new output

To trigger lost events, it was easier to use infer (refer #63)

c\@c : ~/g/runtime_events_tools$ dune exec olly -- gc-stats  "../infer/infer/bin/infer analyze --no-progress-bar --no-report --multicore -j $(getconf _NPROCESSORS_ONLN)"
....
[ring_id=3] Lost 302568 events
[ring_id=4] Lost 141949 events
[ring_id=5] Lost 47645 events
Skipped large procedure (__infer_globals_initializer_ecp_nistz256_precomputed, size:18951) in pulse.

Execution times:
Wall time (s):	130.00
CPU time (s):	1777.23
GC time (s):	669.68
GC overhead (% of CPU time):	37.68%

Per domain stats:
Domain	Wall(s)	GC(s)	GC(%)
0	130.00		28.30	21.77
1	126.48		51.38	40.62
2	127.16		57.20	44.98
3	126.48		50.73	40.11
4	126.47		42.72	33.78
5	126.47		50.86	40.21
6	126.47		50.90	40.24
7	128.89		51.24	39.76
8	126.47		51.00	40.32
9	126.47		32.81	25.94
10	126.47		50.74	40.12
11	126.47		50.50	39.93
12	126.46		50.70	40.09
13	126.46		50.62	40.03

GC latency profile:
#[Mean (ms):	2.27,	Stddev (ms):	74.77]
#[Min (ms):	0.00,	max (ms):	5783.95]

Percentile 	Latency (ms)
25.0000 	0.00
50.0000 	0.01
60.0000 	1.03
70.0000 	1.37
75.0000 	1.61
80.0000 	1.84
85.0000 	2.13
90.0000 	2.65
95.0000 	3.90
96.0000 	4.20
97.0000 	4.63
98.0000 	5.23
99.0000 	7.43
99.9000 	35.78
99.9900 	4840.23
99.9990 	5783.95
99.9999 	5783.95
100.0000 	5783.95

Warning: Results may be inaccurate since some runtime events were lost before they could be read.
Refer to 'Missed events' section in repo README

c\@c: ~/g/runtime_events_tools$ dune exec olly -- trace olly.trace "../infer/infer/bin/infer analyze --no-progress-bar --no-report --multicore -j $(getconf _NPROCESSORS_ONLN)"
....
[ring_id=2] Lost 653481 events
[ring_id=3] Lost 557918 events
Skipped large procedure (__infer_globals_initializer_ecp_nistz256_precomputed, size:18951) in pulse.

Warning: Results may be inaccurate since some runtime events were lost before they could be read.
Refer to 'Missed events' section in repo README

c\@c: ~/g/runtime_events_tools$ 

[curche]

Rebased changes on top of package updates. CI looks ok now