I'll address these one by one.
Should shoulder harnesses be installed, at least to enable passengers to decide to use them or not?
No. They would not serve a significant purpose to improve safety in those seats, and they would add weight and increase maintenance costs for airlines. Keep reading, and this will come to light collectively with my full answer.
Why would shoulder harnesses be obligated for flight attendants in their jump seats, but not for passengers?
Those seats aren't lined up in rows, therefore there is no seat in front of them that they would hit and be stopped by. Think of it like a school bus with no seat belts; the passengers will be stopped by the seat in front of them in a low-force crash, but the bus driver wears a seat belt because there is a steering wheel and windshield in front of them, not a padded seat.
How is 4 true? (It is superfluous in a crash when compared to the same safety from a lap belt)
Essentially, if there is a crash that will kill people, the shoulder harness will not be the thing that makes a difference. The lap belt isn't really even there for crashes. Extreme turbulence could throw someone up out of a seat; a lap belt is sufficient to prevent this.
A shoulder strap is designed to reduce impact with an object on board a vehicle. A car, we can safely assume, will remain in one piece if it hits a concrete wall. Airplanes do not have structures designed to withstand impacts like that. If a plane disintegrates in a crash, the passenger, packaged along with the lap belt, shoulder strap, and all, will come to impact the external object directly due to structural failure. No seat belt can save you from a total structural failure.
Do the safety benefits not outweigh the costs?
Since there really is no safety benefit whatsoever in adding them in the seats that don't already have them, no. It would be security theater that costs airlines a lot of money.
Notice I highlighted the low-force impact statement up above in regards to buses. Let's say there is a low-force impact in a plane such as a taxiing crash that jars everyone on board. The passengers will hit the seats in front of them, not that the seats are even necessary to prevent injury. However the pilots, on the other hand, have a lot of hard instrumentation in front of them such as throttle quadrants, and other things you don't want accidentally shoved forward during a ground crash.
It appears that very little actual research has been done on this subject. For example, this European Transportation Safety Council report states,
Many proposed safety measures require further research before their benefit to cabin
safety and the optimal design can be firmly established. In most cases, the availability
of accurate, validated analytical models is indispensable.
It also states,
It must be emphasised that these [death count] figures are at best estimates, since insufficient detailed
accident information is available.
A fundamental limitation to this process, however, is the lack of adequate accident
information for a sufficient number of accidents to allow full cost benefit analyses to be
performed. The absence in many accident investigations of detailed information on
injury mechanisms and cause of death makes the precise estimation of the potential
benefits of any one measure very difficult.
The Cherry study (1995)*
, one of few studies which have performed a detailed trade-off
concerning the full range of possible survivability measures, provides insight into
available options. However, the small number of accidents available for investigation
in such studies means that the basis for determining priorities still relies heavily upon
best expert judgement rather than a truly numerically-specific approach.
Three-point lap and shoulder harnesses:
As mentioned above, passengers who are restrained by a lap belt only and who do not
take up a brace position prior to impact, are likely to suffer serious injuries due to the
flailing motion of the upper body. As with rearward facing seats, the provision of
three-point shoulder harness restraint systems would prevent this situation. If all
passenger assumed the brace position prior to impact, the additional benefits of a threepoint
shoulder harness would be small. In reality, however, for a variety of reasons,
occupants generally do not assume a proper brace position, so a three-point lap and
shoulder harness would be likely [to] improve occupant protection substantially.
So, after reading this document, I did not find any numerical evidence contained within to support its claims, and it even states that it has very little research to go on and appears to be written in an almost purely speculative and subjective manner, which leads me back to the original point above...
A shoulder strap will not save you from a structural failure, and if the plane bumps into something while rolling wheels-down on the ground, the seat in front of the passenger will absorb their impact.