Only commit memory on windows when initially claiming the frame

[benjaminwinger]

This fixes a fairly significant performance issue on Windows.

I was less familiar with our buffer manager when initially getting it to work on Windows. There's no need to ensure the memory is committed every time we call getFrame.

It was also necessary to skip evicted pages in updateFrameIfPageIsInFrameWithoutLock as it would otherwise segfault from trying to access uncommitted memory (it's not necessary to update them if they're evicted anyway).

[github-actions]

Benchmark Result

Master commit hash: 670db898a2ccb81823d1e444b349eec13183ceb3
Branch commit hash: 50c464588bdccc49b7a98246132e2cd99928cfe7

Query Group	Query Name	Mean Time - Commit (ms)	Mean Time - Master (ms)	Diff
aggregation	q24	686.26	685.54	0.72 (0.11%)
aggregation	q28	10846.82	11391.03	-544.22 (-4.78%)
filter	q14	160.82	160.94	-0.12 (-0.07%)
filter	q15	162.46	162.29	0.17 (0.10%)
filter	q16	342.82	334.05	8.77 (2.63%)
filter	q17	480.33	480.75	-0.41 (-0.09%)
filter	q18	1926.79	1954.48	-27.68 (-1.42%)
fixed_size_expr_evaluator	q07	571.97	571.37	0.59 (0.10%)
fixed_size_expr_evaluator	q08	786.82	788.84	-2.02 (-0.26%)
fixed_size_expr_evaluator	q09	793.36	781.47	11.89 (1.52%)
fixed_size_expr_evaluator	q10	273.15	272.70	0.44 (0.16%)
fixed_size_expr_evaluator	q11	270.88	268.04	2.83 (1.06%)
fixed_size_expr_evaluator	q12	268.30	267.62	0.67 (0.25%)
fixed_size_expr_evaluator	q13	1497.07	1507.57	-10.50 (-0.70%)
fixed_size_seq_scan	q23	152.73	156.53	-3.80 (-2.43%)
join	q31	49.38	46.95	2.43 (5.18%)
ldbc_snb_ic	q35	3679.68	3609.70	69.98 (1.94%)
ldbc_snb_ic	q36	132.34	125.28	7.06 (5.63%)
ldbc_snb_is	q32	10.28	10.74	-0.46 (-4.32%)
ldbc_snb_is	q33	98.71	96.46	2.25 (2.33%)
ldbc_snb_is	q34	98.81	96.66	2.15 (2.22%)
multi-rel	multi-rel-large-scan	3007.05	2905.26	101.80 (3.50%)
multi-rel	multi-rel-lookup	82.91	78.49	4.42 (5.63%)
multi-rel	multi-rel-small-scan	60.51	69.52	-9.01 (-12.96%)
order_by	q25	164.52	164.76	-0.25 (-0.15%)
order_by	q26	478.12	495.45	-17.33 (-3.50%)
order_by	q27	1427.69	1432.58	-4.89 (-0.34%)
scan_after_filter	q01	205.16	205.47	-0.31 (-0.15%)
scan_after_filter	q02	195.42	197.34	-1.92 (-0.97%)
shortest_path_ldbc100	q39	86.13	82.75	3.38 (4.09%)
var_size_expr_evaluator	q03	2101.23	2094.28	6.95 (0.33%)
var_size_expr_evaluator	q04	2220.60	2305.32	-84.72 (-3.68%)
var_size_expr_evaluator	q05	2699.24	2572.47	126.77 (4.93%)
var_size_expr_evaluator	q06	1349.95	1363.64	-13.70 (-1.00%)
var_size_seq_scan	q19	1486.87	1492.22	-5.35 (-0.36%)
var_size_seq_scan	q20	3211.29	3203.26	8.04 (0.25%)
var_size_seq_scan	q21	2452.62	2546.02	-93.40 (-3.67%)
var_size_seq_scan	q22	134.82	141.06	-6.24 (-4.43%)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.13%. Comparing base (13b20ef) to head (fc621bd).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4047      +/-   ##
==========================================
- Coverage   85.13%   85.13%   -0.01%     
==========================================
  Files        1297     1297              
  Lines       50599    50599              
  Branches     6978     6979       +1     
==========================================
- Hits        43079    43076       -3     
- Misses       7379     7382       +3     
  Partials      141      141              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.