Changes

bq27xxx_battery.c: revert 61393cf

bq27xxx_battery.c: use cached flags
The flags were just read by bq27xxx_battery_update(), no need to read
them again.
This may go to upstream.

omap3-n900.dts: remove rx51-battery
N900 has bq27200 chip, which provides much better functionality when
exposing battery properties.
No need to confuse userspace with two battery devices exposed by the
kernel at the same time.

Nokia N900: increase charge current limit to 950mA
That was default in Maemo Fremantle

bq27xxx_battery.c: change sense resistor value
Revert commit f4da8deb.
After some extensive testing I found that initial value of 20 mOhm was
more appropriate than 30 mOhm. The tests:
* Using USB ammeter while charging, I found that measured
 current, exposed by bq27200 driver with sense resistor
 assumed to be 30 mOhm, was much lower than one displayed
 by ammeter. Which shouldn't be so, because of high
 efficiency of voltage converter used in N900.
* I used UT61E to test battery current when charger was
 connected. According to these measurements with relatively
 big currents, the values calculated with 20 mOhm resistor
 assumption were more accurate comparing to 30 mOhm.
* I disposed a battery that was used in N900, soldered wires
 to its terminals and did capacity test with Opus BT-C3100
 with 200mA current. Results:
 - Full charge value by Opus BT-C3100:    1197mAh
 - Full charge value by bq27200 @ 30mOhm: 920mAh
It was done several calibration loops to the battery while Maemo Leste
was running. According to my observations N900 was powered off because
of undervoltage while there was some charge remaining. So 'charge_full'
of that battery should be considered less than 920mAh, if using 30 mOhm
sense resistor value for calculations. Which is not the case if we look
at the measured capacity @ Opus BT-C3100.
According to the tests above, the current value I measured before, when
setting the sense resistor value to 30 mOhm, should not be considered
indicative.

pvr: Add cpufreq transitions postchange guard
It is possible that vdd2_pre_post_func is called for the first time for 
CPUFREQ_POSTCHANGE if we are in the middle of frequency transition when
the module is loaded. This leads to BUG() in pvr_dev_unlock() as there
is no lock held by this time.
Fix that by making sure that CPUFREQ_POSTCHANGE event is processed only 
when there was CPUFREQ_PRECHANGE event before it.
Signed-off-by: Ivaylo Dimitrov <ivo.g.dimitrov.75@gmail.com>

rx51_defconfig: compile RTC module to the kernel
This prevents wrong timestamps on early kernel boot time and removes the
need to have a service/udev rule running 'hwclock --hctosys'.
Without this change, kernel options RTC_HCTOSYS=y and 
RTC_HCTOSYS_DEVICE=rtc0 have no effect except errors in the logs.

rx51_defconfig: disable watchdog nowayout
Allow device to properly shut down and reboot

WIP: gpu/pvr: properly round clock rate
Patch send in by freemangordon, not sure if it should be committed like 
this, hence the WIP.