ppa(LRQ): optimise LRQ entries for ppa

[jin120811]

Previously, due to LRQ can't backpressure, we set LoadQueueReplaySize no smaller than VirtualLoadQueueSize to ensure uop can always enter LRQ. But LRQ is very redundant, so it cause an increase in area and power.

Now, in order to optimize power and area, we need to reduce LRQ redundant entries.

We should resolve this constraint at first.
Use "lqThreshold" to control the execution of load instructions. When the free entries in LRQ reaches the "lqThreshold", we backpressure to make IssueQueue send the oldest uop to LDU.

And then, reduce LoadQueueReplaySize from 72 to 48.

[XiangShanRobot]

[Generated by IPC robot]
commit: 6ee1534

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
6ee1534	1.803	0.441	2.642	1.227	2.154	2.142	2.358	0.938	1.400	1.961	3.104	2.560	2.274	3.293

master branch:

commit	astar	copy_and_run	coremark	gcc	gromacs	hmmer-Vector	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
517c737			2.636			1.501			0.939	1.401					3.303
800b4e0	1.819	0.442	2.636	1.240	2.143		2.140	2.361	0.939	1.405	1.984	3.113	2.565	2.271	3.303
e836c77	1.819	0.442	2.636	1.230	2.143		2.140	2.361	0.939	1.408	1.984	3.113	2.565	2.271	3.303
25a80bc	1.819	0.442	2.636	1.226	2.143		2.140	2.361	0.939	1.408	1.984	3.113	2.565	2.271	3.303
bcdee82	1.823	0.442	2.639	1.230	2.159		2.145	2.361	0.939	1.401	1.931	3.107	2.559	2.272	3.308
b7a6349	1.823	0.442	2.640	1.233	2.159		2.145	2.361	0.939	1.401	1.931	3.107	2.559	2.272	3.308
e9cac66	1.823	0.442	2.639	1.220	2.159		2.145	2.361	0.939	1.401	1.931	3.107	2.559	2.272	3.308
0b4afd3	1.823	0.442	2.639	1.225	2.159		2.145	2.361	0.939	1.405	1.931	3.107	2.559	2.272	3.308